Professional Documents
Culture Documents
Author(s)
Citation
Issue Date
2013-03-25
DOI
Doc URL
http://hdl.handle.net/2115/52995
Right
Type
theses (doctoral)
Additional
Information
File
Information
DoctoralThesis
GeostatisticalReservoirModelingof
TrendingHeterogeneity
SpecifiedinFocusedRechargeZone:
ACaseStudyofToyohiraRiverAlluvialFan,Sapporo,Japan
By
YoshitakaSAKATA
SubmittedfortheDegreeofDoctorofPhilosophy
DepartmentofNaturalHistoryofSciences
GraduateSchoolofScience
HokkaidoUniversity,Sapporo,Japan
February,2013
Table of Contents
Abstract ........................................................................................................................ i
Acknowledgements .................................................................................................. iv
Figures ....................................................................................................................... vi
Tables ......................................................................................................................... ix
Abbreviations ............................................................................................................. x
1
Introduction.......................................................................................................... 1
1.1
1.2
1.3
1.4
1.5
1.5.1
1.5.2
Subsurface hydrogeology.................................................................................. 22
1.5.3
1.6
Summary...................................................................................................................... 28
Introduction 29
2.2
2.2.1
2.2.2
2.2.3
Cross-validation ........................................................................................ 40
2.3
2.3.1
Data filtering.............................................................................................. 40
2.3.2
2.3.3
VHG map.................................................................................................. 47
2.4
Introduction 55
3.2
3.2.1
3.2.2
3.2.3
Observation transects............................................................................... 62
3.2.4
3.3
3.3.1
3.3.2
3.3.3
3.3.4
3.4
Conclusions 80
Introduction 81
4.2
4.2.1
4.2.2
4.2.3
4.3
4.3.1
4.3.2
4.3.3
4.3.4
4.3.5
4.4
Conclusions 110
Introduction 112
5.2
5.2.1
5.2.2
5.2.3
5.2.4
5.2.5
5.2.6
5.3
5.3.1
Observation results..................................................................................131
5.3.2
5.3.3
5.3.4
5.4
Conclusions 150
6.2
Appendix A 159
Appendix B 185
Appendix C 199
References 227
Abstract
Coarse alluvial deposits are increasingly important as water reservoirs, especially in arid and
semi-arid regions. Coarse alluvial deposits consist mainly of poorly sorted sand and gravel, and
the geologic heterogeneity is generally large and trending as a result of depositional and
post-depositional processes. Geostatistical approaches in groundwater reservoir modeling are
various, but are often based on the assumption of stationarity. This assumption is not necessarily
valid in coarse alluvial deposits, and a standard approach in trending heterogeneity is to separate
the target variable (e.g., hydraulic conductivity) into a global trend component and a residual
component. However, the trend component has rarely been determined because of scarcity and
uncertainty in measurements, especially in deep parts of deposits.
This dissertation primarily covers a focused recharge zone around and beneath a losing river
flowing on coarse alluvial deposits. The zone is just one part of a basin, and is thus limited in
area. However, the complex system of groundwater flow and solute transport is directly reflected
in the large and trending heterogeneity. First, surface water infiltration from the riverbed induces
downward groundwater flows. Next, lateral flow components are gradually added, and the
magnitudes are changed depending on geologic heterogeneity. In particular, connected voids in
gravel deposits give rise to preferential groundwater flows. If the depth dependence of
permeability exhibits, for example, a coarsening-upward trend, the groundwater flows
accumulate in the shallower parts. As a result of the complex flow system, solute concentration
and water temperature are characterized specifically in space and time. Thus, detailed research in
a limited zone will reveal the trending heterogeneity in coarse alluvial deposits, and will allow
assessment of its importance in groundwater modeling.
Targeting the Toyohira River alluvial fan (Sapporo, Hokkaido, Japan), this dissertation
presents well data analysis, flow measurements, permeability modeling, and stochastic simulation.
First, vertical hydraulic gradient in the fan is mapped to elucidate the three-dimensional
groundwater flow system in the fan, and to determine the focused recharge zone along the
Toyohira River. The wide range of uncertainty in available well data is addressed by combining a
filtering process, kriging interpolation, and cross-validation. Consequently, the reasonably
accurate maps show that there are downward groundwater flows in the fan and a negative peak in
ii
hand, a block-averaged model of trending heterogeneity also yields optimal realizations that are
slightly inferior to those of a high-resolution model in terms of agreement between calculations
and observations. However, the number of optimal realizations is larger than that of a
high-resolution model, indicating the applicability of block averaging to a larger basin model.
A current avenue of the research is to expand the two-dimensional high-resolution model to a
three-dimensional basin model, not only for the Toyohira River alluvial fan but also for other
coarse alluvial aquifers. Lastly, this dissertation discusses several problems to be addressed with
the advanced model in order to determine hydrogeologic sequences, diffuse recharge,
urbanization effects, and upscaling procedure.
iii
Acknowledgements
There are a number of people who deserve recognition for contributing to this dissertation. In
particular, my advisor, Prof. Ryuji Ikeda, offered wisdom and patience that were singularly
responsible for helping me complete this dissertation. First, he graciously considered my
admission to the doctoral course, even though I had just finished my undergraduate degree and
had no scientific publications under review. During my doctoral studies, Prof. Ikeda consistently
broadened my knowledge, brought out the best in my ideas, and inspired my research directions.
Most noteworthy was his encouragement, which made a deep impression on me: You have to
challenge yourself to do true scientific research, something that any scientistnot just
hydrologistswill recognize as significant work.
I am deeply appreciative of Assoc. Prof. Kazuhisa Chikita, who always pointed out
illuminating perspectives in our hydrology seminars. The two other members of my thesis
committee were Prof. Kosuke Heki and Prof. Noriyuki Suzuki, whose comments and suggestions
led to many scientific improvements. I also wish to thank the many undergraduate and graduate
students in my research group, especially Nobuto Nakatsukasa, Tomoyuki Wada, Wataru Iwasaka
and three BangladeshiMohammad Tazul Islam, Abdullah Al Mamun and Mohammad Motaleb
Hossain. You helped make my back to school life enjoyable through your friendship.
I am very grateful to the many people outside of my university who lent their support,
including Prof. Satoshi Okamura of the Hokkaido University of Education and Dr. Tsumoru
Sagayama of the Geological Survey of Hokkaido, who performed volcanic ash analysis and diatom
analysis on gravel cores. Mr. Daisuke Nagaoka and Dr. Kenji Kizaki provided helpful comments
about the fans geology, which I received with gratitude. I thank H. Fukami of the Geological
Survey of Hokkaido and M. Ishizuka of Aqua Geo Techno, for providing private well data. I would
also like to give special thanks to Takayuki Hosono and his staff at ACE Shisui, whose excellent
core-tube sampler offered me a starting point for my dissertation, and to Kazunori Ito, who
greatly helped me in my previous research on the groundwater of the Toyohira River alluvial fan.
Yoshikatsu Tezuka and Ayano Matsuura of Geo-tech worked hard to perform laboratory soil tests,
and Kojin Nishino supported me in taking well measurements in all seasons. Yoshiyuki Tada and
Ryosuke Kobayashi of the Hokkaido Kasen-Kaihatsu graciously guided synoptic surveys of the
iv
Toyohira River. Takahiro Jyonishi of Hokkai-Suiko Consultants also performed valuable ADCP
measurements in the Toyohira River. I also gratefully acknowledge the Hokkaido Regional
Development Bureau, Hokkaido Prefecture, and Sapporo City for providing valuable data and
reports.
This work would not have been possible without understanding and financial support from my
employer, Docon Co., Ltd., Sapporo, Japan. My managers Hidetoshi Kon, Jun Yamazaki,
Masanori Araki, Kei Tomioka, Satoshi Kumakura, and Yoshiya Hitomi patiently allowed me to
devote three years to my studies. My co-workers deserve special thanks, especially Yoshiaki
Tago and Kazuki Okada, who handled office and field work, much of which should have been
done by me instead. I am appreciative of Sayuri Watanabe and Yoko Yamamoto for diligently
working to input an enormous amount of data, for making my figures and table beautiful, and for
offering helpful revisions throughout this thesis. Also, Akinari Honda of Gm-Labo and Hideki
Ohtake of the Geo-tech gave helpful assistance in computer programming.
My wife, Riko, has offered tremendous support and endless patience throughout my work on
this dissertation. I also wish to apologize to my lovely daughters, Hiyori and RinI did not
spend as much time with you as I should have during your precious childhood. I dedicate this
thesis to my family; I hope it will make you proud.
Figures
Figure 1.1 Study context ...........................................................................................5
Figure 1.2 Connection of groundwater flow and solute transport models .................6
Figure 1.3 Illustrative trending heterogeneity in an alluvial fan ...............................10
Figure 1.4 Schematic system of groundwater flow and solute transport in a focused
recharge zone..................................................................................................15
Figure 1.5 Study site: the Toyohira River alluvial fan...............................................18
Figure 1.6 Photographic scenes of the Toyohira River............................................19
Figure 1.7 A birds eye view of the Toyohira Alluvial fan, indicating its topographic
features............................................................................................................21
Figure 1.8 Geologic profiles along a longitudinal direction in Sapporo City ............24
Figure 1.9 Photographic outcrops in the Toyohira River..........................................25
Figure 2.1 Location maps and topographic features of the study area. ..................33
Figure 2.2 Photos of water level measurement in observation wells.......................35
Figure 2.3 Histograms of well depths and screen lengths.......................................36
Figure 2.4 Definition of inputs for each well (piezometer) .......................................39
Figure 2.5 Long-term variations in annual mean water levels and box plots of
groundwater levels in the observation wells.....................................................41
Figure 2.6 Omnidirectional variogram and fitted spherical model for residuals of GTEi,
.........................................................................................................................43
Figure 2.7 Cross-validation results of GTE above sea level....................................44
Figure 2.8 Maps of (a) GTE and (b) GTD in the Toyohira River alluvial fan ............45
Figure 2.9 Omnidirectional variogram and fitted spherical model for VHGj in three
case .................................................................................................................46
Figure 2.10 Maps of VHG and VHG variance in the Toyohira River alluvial fan......51
Figure 2.11 Maps of VHG and VHG variance in the Toyohira River alluvial fan ......52
Figure 2.12 Maps of VHG and VHG variance in the Toyohira River alluvial fan......53
Figure 3.1 Current meters and Handheld ADV Diagram of the FlowTracker probe.60
Figure 3.2 Photograph of handheld acoustic Doppler velocimeter Flow Tracker ..60
Figure 3.3 Diagram of 5-point mean depth measurements .....................................61
vi
Figure 3.4 Map of Toyohira alluvial fan and gaging stations for synoptic surveys ...63
Figure 3.5 Photos of current propeller Sanei Type 1-P .........................................64
Figure 3.6 Photos of acoustic Doppler current profiler, Stream Pro ......................65
Figure 3.7 Example of measurement results of the Kariki Station...........................68
Figure 3.8 Synoptic survey results ..........................................................................70
Figure 3.9 Relation between river discharge and leakage ......................................73
Figure 3.10 Example of measurement results obtained using ADCP......................78
Figure 3.11 Comparison of ADV and ADCP results.................................................79
Figure 4.1 Schematic of improved double core-tube sampler used in this study ....85
Figure 4.2 Undisturbed cores obtained at BW03 ....................................................86
Figure 4.3 Location maps and geologic features of the study area.........................87
Figure 4.4 Relation between previous results of slug test and mid-depths of the test
screens ............................................................................................................89
Figure 4.5 Classification of gravel core according to matrix packing level ..............90
Figure 4.6 Box and whisker plots showing the vertical statistics of length fractions of
low packing parts .............................................................................................91
Figure 4.7 Grain size distribution curves of undisturbed cores................................95
Figure 4.8 Relations between grain size and hydraulic conductivity obtained by slug
tests. ................................................................................................................96
Figure 4.9 Comparison in double logarithmic scale between hydraulic conductivities
measured by slug test and estimated by Eq. (4.9)...........................................98
Figure 4.10 Curves relating K calculated by Eq. (4.8) and effective grain size
diameter for several length fractions of low packing part ...............................100
Figure 4.11 Core property and hydraulic conductivity results................................104
Figure 4.12 Scatter plot of K values with associated regression lines...................105
Figure 4.13 Longitudinal cross section showing hydrogeologic boundaries..........109
Figure 5.1 Configuration of the off-stream well transect on the Toyohira River ..... 116
Figure 5.2 Explanatary vertical cross-section of the offset-stream transects ........ 117
Figure 5.3 HOBO pressure transducer and HOBO thermal sensor ...................... 117
Figure 5.4 Two dimensional models of high resolution and of block average .......124
Figure 5.5 Schematic River boundary...................................................................126
vii
Figure 5.6 Hourly water-level and temperature variations in the river and observation
wells during the analysis periods ...................................................................132
Figure 5.7 Contour maps of water level and temperature .....................................133
Figure 5.8 Seasonal temperature profiles in the deep well ...................................134
Figure 5.9 Relation between hydraulic conductivity at the ground surface and
coordinates ....................................................................................................136
Figure 5.10 Variograms of spatial residuals of hydraulic conductivity ...................137
Figure 5.11 Examples of two-dimensional stochastic realizations of logK ............140
Figure 5.12 Calculations vs measurements in hydraulic head, river leakage, and
groundwater temperature in the optimum realizations ...................................144
Figure 5.13 Calculations vs measurements between original and block averaged
models ...........................................................................................................146
Figure 6.1 Conceptualization of three-dimensional basin model of alluvial fan .....154
Figure 6.2 Schematic diagram illustrating the scale-up procedure and the upscaling
analysis..........................................................................................................158
viii
Tables
Table 2.1 Long term observation wells in the Toyohira River alluvial fan.................34
Table 2.2 Results of Mann-Kendall test and Sens slope estimation .......................42
Table 2.3 Cross validation results for VHG mapping ...............................................48
Table 3.1 Measurement results by synoptic surveys using ADV .............................67
Table 3.2 River leakage calculated using the stream water budget method ...........74
Table 3.3 Comparison between results of synoptic surveys by ADV and conventional
measurements by propeller .............................................................................76
Table 4.1 Example cores of matrix packing level I to IV ..........................................91
Table 4.2 Results of parameter optimization using undisturbed cores and slug tests
.........................................................................................................................97
Table 4.3 Analysis data and optimization results to determine the relation between
slug tests and gravel cores ..............................................................................99
Table 5.1 Observation wells in the off-stream transect.......................................... 115
Table 5.2 In-situ measurements of hydraulic conductivity .....................................121
Table 5.3 Input model and parameter values ........................................................127
Table 5.4 Description of the simulation cases .......................................................130
Table 5.5 Individual numbers and total number of optimum realizations ...............141
ix
Abbreviations
a: practical range of variogram
ADCP: Acoustic Doppler current profiler
ADV: Acoustic Doppler velocimeter
B: Nugget of variogram
C: Sill of variogram
C(xi, xj): Covariance between xi and xj
CV: Cross validation
d: Diameter of grain size
de: effective grain diameter (d20 in this study)
D: Day
DEM: Digital elevation model
E[x]: Statistical expectation of x
GLS: generalized least squares
GTE: Groundwater table elevation in meter above sea level
GTD: Groundwater table depth from ground surface
h: Hydraulic head
h: the vector between measurement points
HRDB: Hokkaido Regional Development Bureau
ISO: International Organization for Standardization
K: Hydraulic conductivity
K: Hydraulic conductivity estimated by permeability model
xi
1 Introduction
1.1
Research motivation
The world faces unprecedented challenges concerning water resources, particularly in arid and
semi-arid regions, which occupy 47.2% of Earths land surface (UNEP 1991). Recent
Intergovernmental Panel on Climate Change estimates (Kundzewicz et al. 2007) indicate that
between 1.4 and 2.1 billion people live in areas experiencing water stress and these numbers are
growing under pressures from population growth and climate change. In such regions,
groundwater resources have immense value in terms of natural storage of water and potential
storage that can be managed.
Coarse alluvial deposits are not major components in stratigraphic records of the land surface
(Rust 1979). However, gravel deposits have generally served as vast reservoirs and significantly
contributed to industry, agriculture, and urbanization in Japan and around the world. In particular,
thick and monotonic gravel sequences are often seen adjacent to high relief terrain in arid and
semi-arid environments, where sparse vegetation and occasional intense rainfall promote
aggradation of coarse deposits. However, the extent of coarse deposits is often limited,
particularly around the foot of mountains, and so the groundwater capacity is not inexhaustible.
Sustainable use of water, especially in urbanized and developing areas, is threatened by excessive
withdrawal, contaminant infiltration, watershed development, climate change, and other factors.
In addition, the interaction between surface water and groundwater reduces the flow rates to the
minimal values required for riparian environments. The permeability of coarse alluvial deposits is
generally high, and zones of dewatering and pollution are rapidly and widely expanding.
Hydrogeologic heterogeneity unpredictably complicates the spatial and temporal range of an
aquifers influence.
Numerous models have been used in practice to interpret actual flows and transports in
groundwater systems. Recently, various models have been developed of parameters such as flow,
solute transport, reactive transport, multiphase flow, and energy flow. A complete groundwater
model that gives predictions in exact agreement with measurements is ideal but never obtainable.
The practical reason is a lack of hydrologic and hydrogeologic information, especially in deep
parts of deposits. For example, hydraulic conductivity K ranges over several orders of magnitude,
even in the same geologic sequence. This is because K depends on various properties of geologic
materials, such as grain size, porosity, sorting, and packing. Geostatistical approaches are
commonly used in many groundwater models to realize spatial variability of K in a target aquifer.
However, the approach is commonly based on an assumption of stationarity, that is, constant
mean and variance in the domain. Moreover, in alluvial deposits, geologic heterogeneity
generally shows horizontal and vertical trends as a result of depositional and post-depositional
processes. A standard geostatistical approach in nonstationary fields is to separate the target
variable (e.g., K) into a global trend component and a residual component; however, the global
trend has rarely been determined because of scarcity and uncertainty in measurements, especially
in deep parts of deposits. The necessity of performing many measurements conflicts with the
advantage of geostatistics.
This dissertation presents research on trending heterogeneity in coarse alluvial deposits and the
development of an approach to modeling their large and trending heterogeneity. For the purpose,
attention is paid to the area of a focused recharge zone. Focused recharge zones are typically
found beneath and around losing rivers as a result of interaction between surface water and
groundwater. The focused recharge zone corresponds to just a part of the regional
three-dimensional groundwater system, but spatial and temporal distributions of hydraulic head
and groundwater temperature are typically characterized by trending heterogeneity. Regarding
the Toyohira River alluvial fan (Sapporo, Hokkaido, Japan), this dissertation covers well data
analysis, flow measurements, permeability modeling, and stochastic simulations. This fan is
taken up as a representative case for this research object because of its large and trending
heterogeneity in gravel deposits.
1.2
Thesis outline
Chapter 1Introduction
This chapter first describes the motivation for this research, followed by the standard
approaches and their problems in groundwater modeling, typically in coarse alluvial deposits.
Next, the aim and scope for addressing the typical problems are pointed out. This chapter further
describes the study site, including its topography, hydrogeology, and previous investigations.
Chapter 2Regional mapping of vertical hydraulic gradient by well data analysis and
kriging interpolation
This chapter describes a mapping method of hydraulic head for delineating the
three-dimensional groundwater flow system in the fan, and for determining the focused recharge
zone. For this purpose, mapping groundwater table elevation and vertical hydraulic gradient is
performed. The mapping process comprises long-term water-level analysis and kriging
interpolation to address the wide range of uncertainty in conventional well data. This content is
included in a published paper entitled Regional mapping of vertical hydraulic gradient using
uncertain well data: a case study of the Toyohira River alluvial fan, Japan (Sakata and Ikeda
2013) in the Journal of Water Resource and Protection, 5(8) in press.
Chapter 3Synoptic survey for quantifying discharge and leakage in gravel-bed river
Chapter 3 presents an improvement in synoptic survey methodology for specifying the focused
recharge zone and for estimating the rate of seepage loss in the zone. This chapter firstly points
out how to reduce the discharge uncertainty in gravel-bed rivers. A high-accuracy current meter is
applied and verticals are arranged in detail. In addition, the efficiency of the proposed method is
indicated as a result of comparisons with other methods.
This content is included in a published paper entitled Quantification of longitudinal river
discharge and leakage in an alluvial Fan by synoptic survey using handheld ADV (Sakata and
Ikeda 2013) in the Journal of Japan Society of Hydrology and Water Resources, 25(2), 89-102.
Chapter 4Realization of heterogeneity in permeability of gravel deposits
In chapter 4, a permeability model of gravel deposits is constructed by using a relatively
undisturbed core. A degree of packing of fine sediments between gravel grains is visually
classified by four packing levels, and a good relation between slug test results and core properties
1.3
The governing equation of groundwater flow is derived by combining the mass conservation
equation and Darcy's law (Appendix I-1). The equation is mathematically represented as a partial
differential equation of hydraulic head. Hydraulic head is the potential of groundwater flow. The
equation contains several parameters, including hydraulic conductivity, specific storage, and
sink/source terms. Solving the equation requires appropriate boundary conditions, such as the
no-flow condition, Dirichlet condition, Neumann condition, and Cauchy condition. In the
transient simulation, initial conditions are also important for obtaining reasonable solutions.
No information about these parameters and conditions is obtainable throughout the entire
model domain, so groundwater models inherently have discrepancies derived from uncertainty. In
most case, it is difficult to construct sufficient models that give calculation results in exact
agreement with measurements in terms with hydraulic head or solute concentration. A
fundamental problem in groundwater modeling is how various parameters are determined in
space and time. This section discusses standard approaches to groundwater modeling in more
detail, and points out their inevitable problems, especially in coarse alluvial deposits. Here, the
term coarse alluvial deposits is used for poorly sorted sand and gravel, typically containing 50%
or more gravel (particles greater than 2 mm in diameter). In fact, the physical characteristics of
the deposits (e.g., grain size distribution and porosity) are dependent on mainly the depositional
system by fluvial or debris flows.
Figure 1.2 Connection of groundwater flow and solute transport models (Carrera and
Mathias 2010)
budget methods are often applied for estimating stream loss (Healy 2010). This research employs
a conventional treatment of focused recharge zone, such that uncertainty in groundwater recharge
is assumed to be negligible and trending heterogeneity (another uncertain parameter) can be
discussed in more detail. Among water budget methods, a synoptic survey of discharge is a
widely used and effective method to determine focused recharge (Woessner 2000; Weight 2008).
However, such surveys have been carried out in only small and gentle rivers due to uncertainty in
discharge measurement. In typical gravel-bed rivers, unsteady turbulent flow and roughness of
the bed boundary significantly affect the application of synoptic surveying in focused recharge
estimation.
Realization of K with large and trending heterogeneity
Hydraulic conductivity K, a proportionality coefficient in Darcys law, generally spans several
orders of magnitude, even in the same geologic setting. The spatial variability of K is related to
various geologic material properties such as grain size, porosity, sorting, and packing. An
application of geostatistical and stochastic methods for groundwater reservoir modeling has
become common practice to address large heterogeneity (Koltermann and Gorelick 1994; Dagan
1997; Jang and Liu 2005; Eaton 2006; Lee et al. 2007). Another zonal and deterministic
approaches have been more popular in practice, and various optimization methods are developed
(e.g., Horino 1992). However, the deterministic approaches are less flexible because only one
optimal realization of K in the target domain is obtained under insufficient measurements of K.
The geostatistical approach is considered to be more effective in groundwater modeling of coarse
alluvial deposits, because coarse alluvial deposits often consist of monotonic gravelly sequences
and have few obvious boundaries, necessitating the zonal approach.
Figure 1.3 shows an illustration of heterogeneity in coarse alluvial deposits, in this case,
alluvial fan gravel deposits. On a basin scale, various sequences are seen that consist of
clast-supported gravel (well sorted), matrix-supported gravel (less sorted), sand, silt and others.
The sequences result from the complex and changing depositional processes. In general, however,
a downward fining trend is generally caused by decreasing flow energy. This trend is called
trending heterogeneity by Freeze and Cherry (1979). Humped heterogeneity is also seen on the
basin scale and indicates that transmissivity peaks in the mid-fan area as a result of sorting
trends; in other words, sheetflood and channel bar deposits in the mid-fan to lower fan are much
better sorted in comparison with the debris and mud flows in the upper fan. In alluvial fan
aquifers, a humped trend is probably more common than a simple downward trend (Cehrs 1979;
Neton et al. 1994).
In the vertical direction, the stratigraphic sequences have large heterogeneity because they are
composed of various hydrofacies due to depositional and post-depositional processes. For
example, coarsening/thickening or fining/thinning sequences are often formed by progradation,
retrogradation, and basin subsidence (Neton 1994). However, these trends are not necessarily
reflected in vertical profiles of hydraulic conductivity K because it is a complex function of
various factors of geologic materials (porosity, packing, sorting, etc.). In addition, no obvious
trend has been found in the profiles of K, owing to a combination of large heterogeneity and
scarce measurements. The post-depositional processes that influence K (e.g., compaction) have
rarely been considered in coarse alluvial deposits. For example, typical disturbed cores of gravel
deposits provide less information about porosity structure related to permeability, and the
physical properties from logging and surveying indicate weaker correlation with K. Consequently,
the vertical profiles have proved especially difficult in alluvial shallow deposits.
Being fundamentally based on an assumption of stationarity, geostatistics lacks versatility in
groundwater reservoir modeling. This assumption is invalidated by trending heterogeneity of
geologic deposits, for example, where K decreases overall as the grain size gradually decreases
with increasing depth. The problem is called a non-stationary problem in geostatistics, and has
been discussed in the literature (e.g., de Marsily 1986; Wackernagel 2003). One simple way to
address non-stationary problem is to separate target attribute (e.g., K in this case) into a global
trend component and a random residual component. The standard method for handling
non-stationarity requires many measurements to reveal the global trend in large variations of K,
which conflicts with the advantage of geostatistics.
Horizontal direction
U
lower fan
Vertical direction
upper fan
mid fan
Proximal
Distal
A
down-fan fining trend
Dominant
(Blissenbach 1954:
Hooke 1967:Bull 1972)
Common
Lower fan
Streamflow/Stream
channel
Sheet flood
Sieve deposition
Debris flow
Rare
Mid fan
Fan head
Sieve deposition
Debris flow
Sheet flood
Streamflow/Stream Sieve deposition
channel
Sheet flood
Debris flow
Streamflow/Stream
channel
B
trending heterogeneity
Fan head
Mid fan
Lower fan
0
C
humped heterogeneity
(after Wilson 1973;Cehrs 1979)
Sourceland
Distance
0
Gms
Fm+C
Facies description
Gm
St and Sh
Figure 1.3 Illustrative trending heterogeneity in an alluvial fan, modified from Neton (1994) and Einsele (2000)
10
11
1.4
The previous section points out several problems in groundwater reservoir modeling of coarse
alluvial deposits. This section describes the aim and scope for addressing the inherent problems.
Scope 1: Research on focused recharge zones
In this dissertation, special attention is paid to the focused recharge zone around the losing
river. A schematic vertical system of groundwater flow and solute transport is shown in Figure
1.4. The water table is smoothly connected between surface water and groundwater, and the
water table forms a mound around the river, indicating seepage loss through the riverbed. The
infiltrated water first moves primarily in the vertical direction, and then its direction varies
widely in the horizontal and vertical directions, depending on hydraulic gradient. In addition, the
magnitudes of individual flow vectors are spatially variable according to K at a given point.
Solute components and surface water heat are also transported from the river to the deeper zone.
As a result of the preferential flows, sharp solute concentration fronts surround the losing river.
Assuming a decreasing trend of K in the aquifer, the flow vectors of groundwater would largely
accumulate in the shallower zone, and the contours of solute concentration are characterized by
this accumulation. Therefore, the focused recharge zone is just a part of the regional
three-dimensional system of groundwater, but the complex flow field is directly reflected in the
geologic heterogeneity. In other words, this research on a focused recharge zone offers an
effective method for revealing the trending heterogeneity in the groundwater flow and solute
transport system.
Scope 2: Field and analysis methods for reduction of data uncertainty
This research also has a typical scope regarding how to reduce uncertainty in field
measurements. Available water levels for head mapping are filtered, interpolated by kriging, and
cross-validated (Chapter 2). A synoptic survey method is developed that employs a
high-accuracy current meter and a detailed arrangement of verticals (Chapter 3). An exponential
model in vertical K profiles of formulated through a moving average method (Chapter 4).
Scope 3: Permeability model of alluvial fan gravel deposits
A number of K values in a vertical profile are needed to reveal a global trend (i.e., depth
dependence) of K in coarse alluvial deposits. A permeability model of gravel deposits is newly
12
established by using relatively undisturbed cores, which provide typical information on porosity
structure in the deep zone. For this purpose, an additional matrix packing level index is proposed,
and length fractions of each packing level are measured in each core. The permeability model of
gravel deposits consists of effective grain size and length fractions, and the constants are
optimally set. The profiles of core properties in the fan-apex, mid-fan, and fan-toe are
transformed to profiles of estimated K. The depth dependence of K is then represented as an
exponential model through regression analysis using the profiles of estimated K.
Scope 4: Stochastic simulation of trending heterogeneity
Residuals of K from the global trend are expected to be stationary. If this assumption is valid
and the variogram of Re is known, multiple realizations of K in the analysis domain can be
produced by geostatistical methods. This study applies sequential Gaussian simulation, which is
among the most practical and widely used geostatistical methods, yielding 100 realizations of K
in each study case. Groundwater flow and heat transport simulation is then performed under each
condition of K, and optimal realizations to satisfy the criterion are extracted from among the
realizations. Two study cases, namely, trending heterogeneity and stationary heterogeneity, are
prepared to assess an importance of modeling trending heterogeneity in the focused recharge
zone. Moreover, other study cases are examined to discuss other uncertain parameters such as
anisotropy of the unit cell, River conditions (vertical hydraulic conductivity in river-bed material
and river width), range of the variogram, thermal dispersivity, and temperature dependence of
fluid properties.
Scope 5: Application of heat tracer and high-resolution model
The previous section describes three obstacles to obtaining robust solutions in solute transport
simulations. The first and second problems are addressed by using heat as a groundwater tracer
by exploiting the similarity between temperature transport and solute transport: compared with
other solute components, heat can be measured with greater ease and accuracy, and the
representative values of thermal parameters are obtained from the literature. The third problem is
also addressed by using a high-resolution grid model. The Pclet number Pe is commonly used as
a criterion for setting grid size:
13
Pe
w c w qL (1.1)
ke
where w is the density of fluid, cw is the specific heat per unit volume of water, q is Darcys
specific discharge, L is a characteristic length, and ke is the effective (bulk) thermal conductivity.
Darcys specific discharge is calculated by multiplying K by hydraulic gradient. Equation (1.1)
indicates that stable solutions require finer grid-spacing as K increases. Coarse alluvial deposits
in particular have extremely high K values, and so a high-resolution grid model of 1 m
grid-spacing is necessary to satisfy the criterion.
The program SEAWAT is utilized for variable density and transient groundwater flow in
porous media. K in a geologic material consists of fluid density and viscosity in addition to
intrinsic permeability, such that K is inherently dependent on groundwater temperature. In most
conventional models, the effect of temperature is ignored. Recent studies, however, suggest the
importance of temperature dependence with respect to solution robustness in groundwater
modeling (Ma and Zheng 2010; Engeler et al. 2011). SEAWAT also calculates groundwater flow
and temperature transport with greater accuracy by using iterative processes.
A block-averaging model consisting of 5 m square cells is also prepared for comparison with
the high-resolution model. Upscaling of a high-resolution model to a basin-scale model is a
long-standing problem in groundwater modeling (Deutsch 2002; Zhang et al. 2006; Zhang et al.
2010). This is because K depends on scale, that is, microscale (e.g., drilling cores; several meters
or less), mesoscale (e.g., facies; 10 to 100 m), and macroscale (basin scale, 100 m or more). The
discrepancy between the different discretized models shows a direction of upscaling in
groundwater modeling of coarse alluvial deposits.
14
Riverbed
sediment
Water table
Profile of K
h2
Potentiometric line
at the aquifer bottom
h1
Actual flow
Contour of temperature
15
1.5
An alluvial fan (Senjyouchi in Japanese) is a body of alluvial material deposited at the foot of a
range of hills or mountains by a stream that debouches from the area undergoing erosion above the
apex. Alluvial fans are commonly found in arid and semi-arid regions with tectonically active
mountains where there is an abundant supply of sediment. Alluvial fans also occur in subtropical,
arctic, alpine, and humid temperate environments around the world (Lecce 1990). For example,
Japan, despite having a humid climate, also has over 200 alluvial fans (Saito 1980). The
following reasons explain fan development in Japan: steep land slopes, mountainous terrain,
tectonic activity, and base rocks of relatively a recent epoch with low resistivity to erosion. Ono
(1990) also documented that alluvial fans in Japan (and in Korea) occurred mainly because of
climate change, particularly during the early half of the glacial age (90,000-40,000 yr B.P.) and the
short time of the Holocene epoch.
The study site is the Toyohira River alluvial fan in Sapporo, Japan (Figure 1.5). Sapporo City is
the largest in the northern part of Japan and has a population exceeding 1.9 million. The Toyohira
River alluvial fan is located at latitude 43N and longitude 141E, where the Toyohira River flows
northward out of Tertiary mountains and debouches on a plain. The fan has a radius of about 7 km
and an area of about 31 km2. The Toyohira River is also about 72.5 km long and has a watershed
area of about 900 km2. In the watershed, the annual mean precipitation between 1981 and 2010 is
1,107 mm. Snow falls on the fan to a total depth of more than 5 m in the winter. The flow gaging
stations are located at the distal part (KP 11.1; Kariki station; Figure 1.6a) and the apex (KP 18.0;
Moiwa station), respectively, where KP denotes the distance in kilometers from the confluence
with the Ishikari River, one of the main rivers of Japan. The median value of daily mean river flow
through the fan are 12.6 m3/s for the period of record 19752010; maximum instantaneous peak
discharge reached 700 m3/s in August 1981. The discharge in the river is continuously controlled
by two dams: the Jyouzankei Dam and the Hoheikyo Dam, which were constructed in 1972 and
1988, respectively. At the upper point 10 km from the fan, water intake is conducted for
electricity generation. The water after the generation flows into the Yamahana River, which meets
the Toyohira River at KP 18.8, where the Toyohira River usually reaches its peak flow rate
16
throughout the fan. Two other tributaries of the Toyohira River flow through the fan: the
Makomanai River and the Shojin River. The flows of both tributaries are too small to affect the
flow of the Toyohira River. In addition, water is continuously pumped from the Toyohira River at
KP 17.4 (no more than about 0.3 m3/s) to maintain the Sousei River, flowing northward through
the city.
Management of the river, especially during low discharge, has recently become problematic
due to increases in water intake and the active interaction between the surface water and
groundwater. In addition, contaminants in some wells such as natural arsenic and artificial
volatile organic compounds have reached concentrations that are higher than stipulated levels
(Sapporo city 2008; 2011). Various groundwater analyses have been conducted to examine the
river management issues, and the results of several of these analyses have obtained fairly good
agreement with at least part of the measured factors (i.e., groundwater heads, the recharge and
water budget or contaminant transport).
17
Sapporo city
Sosei River
Hokkaido
N45
E140
A'
E130
Toyohira River
Hokkaido
University
Flo
N40
11
12
Sapporo JR station
N35
13
14
N30
Holocene Surface
15
Kamokamo River
16
Outflow
N4320
Inflo
E14130
Ishikari
River
Ishikari Bay
Tobetsu
River
Toyohira
Alluvial Fan
Hoheikyo-Dam
17
Yubari
River
Asari Mt
Pleistocene
Chitose
River
Moiwa Mt
Shojin River
18
Moiwa Mountain
Yamahana River
19
Toyohira River
Muine Mt
Jozankei-Dam
Kitanosawa River
Koizari Mt
A
20
Shikotsu-ko
Makomanai River
N4240
18
Figure 1.5 Study site: the Toyohira River alluvial fan. The aerial photograph was taken on 3-4 June 2004 by Sin Engineering Consultant, Co., Ltd. The
topographic boundaries of the Holocene and Pleistocene fans are taken from an existing geologic map (Osanai et al. 1974; Ishida et al. 1980)
Figure 1.6 Photographic scenes of the Toyohira River: a) at the apex (KP 20.4)
and b) at the Kariki gaging station (KP 11.1)
19
1.5.1
The fan is surrounded by Tertiary volcanic mountains (the Sapporo southwest mountains) to the
west, the Tukisamu hills to the east, and the lowland to the north (Figure 1.7). The fan is a
compound fan comprising a western Holocene fan at 7010 m asl (above mean sea level) and an
eastern Pleistocene fan at 9020 m asl. The Holocene and Pleistocene fans are respectively called
Sapporo-men and Hiragichi-men in Japanese. The fans are topographically different in mean
inclination of the fan surface; 9/1,000 in the Pleistocene fan and 7.5/1,000 in the Holocene fan. The
topographic boundary between the surfaces is recognized as an abrupt cliff of several meters in
height along the east bank of the river upstream from KP 15. Daimaru (1989; 2003) suggested that
the Toyohira River alluvial fan developed as follows: (1) the Pleistocene fan was formed when the
sea level was lower in the last glacial period; (2) as the sea level rose in the early Holocene, the
Toyohira River shifted toward the west, crossing through the Pleistocene fan, and gradually formed
the Holocene fan; (3) in the late Holocene, the river shifted its channel eastward as sediments were
deposited. Lobes A, B, and C in the fan-toe of the Holocene fan indicate the eastward transition
of the river channel, and lobe D corresponds to the present course of the river. Recently, Nagaoka
et al. (2008) suggested that the Hiragishi surface formed by deposition from the earlier age, namely,
the last glacial age, because the upper fluvial terraces including the Shikotsu volcanic deposits at
41,000 yr B.P. (Hu et al. 2001) were eroded by the river.
20
Elevation (m asl)
500
250
200
150
100
80
20
0
Holocene fan
(Sapporo Surface)
Moiwa Mountain
(Eastern edge of the Sapporo
South-west mountains)
Low land
Sosei River
C
D
A
Pleistocene fan
(Hiragishi Surface)
Yamahana
River
Shojin River
Kitanosawa River
Toyohira River
Figure 1.7 A birds eye view of the Toyohira Alluvial fan, indicating its topographic features
21
1.5.2
Subsurface hydrogeology
Thousands of drillings for civil development have been performed in the Toyohira River
alluvial fan, and the fans subsurface hydrogeology are described in the literature by using these
geologic data (e.g., Kato et al. 1995; Oka 2005; Sagayama et al. 2007; Hu et al. 2010). Figure 1.8
shows two vertical cross sections along the longitudinal direction (see
Figure 1.5); the upper panel is the large-scale cross section made by Oka (2005), and the lower
panel is the small-scale cross section prepared for this dissertation. The bedrock distributed
around the alluvial fan is Neogene sedimentary rocks and volcanic rocks (andesitic lava,
pyroclastic rocks, etc.) from the Neogene and Quaternary. The Neogene sedimentary rocks are
classified into the Otarunai-gawa formation and the Nishino formation. The basement strata
appear just beneath the riverbed at the uppermost of the fan (KP 1820; Figure 1.9a). The fan
basement suddenly inclines northward to a depth of several hundreds of meters. The inclination is
caused by a tectonic process (the Ishikari Depression) and is a main factor in the fan development
incorporating a large amount of debris from the upper mountains. As a result of this depression, in
the mid to the distal parts, the bedrock is covered by Pliocene to early Pleistocene deposits called
as Zaimoku-zawa formation. Over almost the entire fan, except for the fan-apex, the thickness of
the Zaimoku-zawa formation is of the order of 102 m, and the maximum thickness is over 800 m
around the fan-toe (see Naebo in Figure 1.8a). The Pleistocene deposits consist of alternating
gravels, sand, silts, clays, and humus. Instead of base rock, Pleistocene marine sediments (50 to
30 m asl) serve as a hydrogeologic basement for the subsurface.
Recently, Hu et al. (2010) divided the upper formations into four aquifers: Pleistocene nos. III
and IV, and Holocene nos. I and II. The Pleistocene no. IV aquifer consists of alternating gravel,
sand, and silt layers tens of meters thick, indicating that the bottom aquifer formed before the fan
developed, that is, in the fluvial or shallow-sea environment. On the other hand, both the
Pleistocene no. III and Holocene no. II aquifers consist mainly of alluvial fan deposits, that is,
poorly sorted sandy gravel with andesitic cobbles (Figure 1.9b). The grain size of the riverbed has
a mean diameter of 50100 mm, a maximum diameter of 100500 mm, and a sand fraction less
than 10% (Hokkaido Regional Development Bureau 2006a). There is an obvious distinction in the
thickness of fan deposits between the Holocene and the Pleistocene fans: the total thickness is
22
over 100 m at the upper middle of Holocene fan, and less than 50 m at the distal part, while the
thickness ranged between only 5 m to 30 m in the Pleistocene fan. However, conventional drilling
cores (largely disturbed) show only the thicknesses of gravel deposits, and due to their large
disturbance, the hydrogeologic details in the gravel sequences are not clear, except for
accompanying thin silt beds. Permeability in the fan deposits is high but greatly variable due to
geologic heterogeneity. Yamaguchi (1965) suggested that a zone of extremely high permeability
was probably located at a depth of ~30 m because the wells screens accumulated in the shallow
depths.
The Holocene no. I aquifer is distributed in only the northern part of the fan (i.e., the fan-toe),
and is geologically and topographically continuous in the lowland. The less permeable aquifer
consists of sandy beds, clayey beds, and peaty beds, which were transported in the late Holocene.
Figure 1.8 indicates potentiometric contours of groundwater head manually plotted by using
head measurements in several piezometric nests. Alluvial fans have steep ground surfaces from
the apex to the distal part, resulting in downslope and downward directional flows (Tolman 1937).
In theory, the topographic-driven flow system is induced as a type of regional three-dimensional
groundwater flow system (Tth 2009). The potentiometric contours in Figure 1.8 show
groundwater flows in the fan are three-dimensional; horizontally from the apex to the distal part
and vertically from the ground surface to the deeper aquifer. The three-dimensional flows of
groundwater occur because of not only the steep land surfaces, but also the seepage loss around
the focused recharge zone.
23
24
Figure 1.9 Photographic outcrops in the Toyohira River: a) tuff breccia of the
Nishino formation (KP21.0); b) gravelly bed indicating imbrication (KP16.5)
25
1.5.3
Alluvial fans are hydrologically and hydrogeologically well-studied in Japan (Kayane and
Yamamoto 1971; Kayane 1991; Huh et al. 2010a). The first study of groundwater in the Toyohira
alluvial fan was probably conducted by Fukutomi (1928). This pioneer produced a
groundwater-table map before World War II by using 3,400 pieces of well-head data collected in
the Holocene fan, and documented the following: 1) the groundwater table was shallower than
the depth of only about 5 m; 2) the table became gradually shallower northward; and 3) the table
crossed the ground surface at an elevation of about 15 m asl, such that several springs called
memu formed along this level.
In the 1960s, groundwater resources in the fan attracted increasing interest with the economic
growth in Japan, and many studies and investigations are performed to reveal the hydrogeologic
structures in the fan, and to assess the capacity of groundwater (Kawata and Obara 1958; Sasson
Keizai Kyougikai 1963; Ozaki et al. 1965; Yamaguchi et al. 1965; Obara 1969). Numerous water
wells were installed at various depths throughout the aquifers, especially no. II and no. III. A total
pumping rate exceeded 100,000 m3/day in 1963, and the pumping rate continued for at least half
a century according to the annual reports compiled by Sapporo City. From 1969 to 1994, the
Nanboku, Tozai, and Toho subway lines were constructed in turn, and heavy pumping for
construction caused significant drawdowns of water levels around the lines. As a result, many
shallow wells stopped pumping and the springs at the distal part dried up (Nakao 1983). After the
construction, the subway lines resulted in drainage, and the groundwater water levels remained
low. Hence, the vulnerability and sustainability of groundwater resources in the fan have been
threatened due to withdrawal and drainage as a result of the citys development. In addition, the
annual average temperature at the Sapporo District Meteorological Observatory suddenly
increased, probably due to the heat island effect, as well as global warming. The rate of increase
is reported to be 2.3 C per 100 yr from 1901 to 2000; for the winter low temperature in particular,
the rate reaches 5.2 C per 100 yr. Climate change might also have significant effects on the
surface water/groundwater budget in the fan (Hu et al. 2010).
The estimation of seepage loss has been a longstanding issue for the management of the river,
especially during the low-flow periods (Hokkaido Regional Development Bureau 2006a), and
26
repeated synoptic surveys of discharge have been carried out (Ozaki et al. 1965; Yanagiya 1995;
Tanaka et al. 2010). However, the estimated values were significantly different due to large
uncertainty of measurement in the turbulent flows.
A variety of groundwater modeling studies has been conducted to address the issues described
above. First, two-dimensional planar models were applied in the 1990s (Furukawa 1997). The
simulation models showed that the groundwater heads were in approximate agreement with the
measurements, and that the seepage loss was estimated at 20 million m3/yr (0.6 m3/s), indicating
the river leakage was the main source of groundwater in the fan. Next in the 2000s,
three-dimensional models were constructed to capture the complexity of the groundwater flow
system (Koizumi et al. 2008; Sakata and Ito 2009). Koizumis model aimed to evaluate the
influence of climate change on the water budget in the fan, while Sakata and Itos previous model
is aimed to reveal the mechanism of seepage loss around the river. In both models, K was
assumed to be homogeneous in each aquifer (i.e., nos. I to IV), and the values were optimally
determined to obtain agreement in hydraulic head between measurements and calculations. The
zonal approach, however, continued to give the discrepancies in seepage loss between calculated
and measured values. Large heterogeneity of K probably influenced the simulation results more
significantly than other uncertain parameters (i.e., diffuse recharge or urbanization effects). The
practical problem is a starting point for this study.
27
1.6
Summary
Coarse alluvial deposits are valuable water reservoirs, especially in arid and semi-arid regions,
but their large and trending heterogeneity is problematic in groundwater modeling. The aim of
this thesis is to reveal the importance of modeling trending heterogeneity in coarse alluvial
deposits. For this purpose, the target area for this research is specified as a focused recharge zone
around a losing river, such that various other uncertain factors (e.g., diffuse recharge) can be
neglected. In the focused recharge zone, this research involves head mapping, synoptic surveying,
permeability modeling, and stochastic simulation. The first aim is reducing the uncertainty in
head mapping, synoptic surveying, and exponential modeling of permeability. The second aim is
to establish a permeability model for gravel deposits by using the matrix packing level index of
relatively undisturbed cores. The third aim is to utilize heat transport simulation to obtain
numerical solutions to solute transport problems. In addition, a high-resolution model is used to
obtain a steady solution to conduction-convection equation. Flow and heat transport coupling
code is utilized, which includes an algorithm for the temperature dependence of fluid to obtain
more rigorous solutions. Sensitivity analysis is conducted to evaluate the importance of trending
heterogeneity in the groundwater flow system, and to assess other uncertain parameters. The
program SEAWAT is utilized for variable density and transient groundwater flow in porous
media. In particular, a block-averaging model is also designed construct an advanced basin model
for future work. A high-resolution grid model is required for the low value of the Pclet number.
The Toyohira River alluvial fan, Sapporo, Japan, was selected as a representative field in this
investigation. The fan has a typical three-dimensional groundwater flow system due to steep
terrain and fan deposits with high permeability. Urbanization complicates the three-dimensional
flow system. The complexity of the groundwater flow system has been indicated in the previous
investigations, and trending heterogeneity in the gravel sequences are required for describing the
flow system in a more quantitative manner.
28
Introduction
Maps of hydraulic head in observation wells are basic elements for planning water resource
development, monitoring and protecting water quality, and conceptualizing groundwater flow and
solute transport. These maps also play a fundamental role in numerical modeling of groundwater
by providing initial conditions and calibration data. In practice, hydraulic head maps closely
correspond to 2D horizontal cross sections: one or several potentiometric contour maps in confined
aquifer(s), and the groundwater table map in an unconfined aquifer. Such conventional 2D maps
are effective for visualizing groundwater flow systems in flat plains or in localized areas where
vertical components are relatively small. On a wider regional scale, topographic and geologic
variations result in a 3D groundwater flow system characterized by vertical flows (Tth 2009).
Measuring the upward and downward flows of groundwater allows for identification of discharge
and recharge areas. In addition, vertical flows beneath bodies of water are interesting for evaluating
the exchange between surface water and groundwater. Also, excessive withdrawal and numerous
understructures in urbanized and developing regions can capture natural discharges for artificial
use elsewhere. Areal information on vertical groundwater flows is no less critical than horizontal
flows for interpreting actual 3D groundwater flow systems.
Vertical hydraulic gradient (VHG) is generally used for delineating the direction and magnitude
of vertical groundwater flows (e.g., Fritz and Mackley 2010). Theoretically, hydraulic gradient
can be calculated by derivation of hydraulic head (Abriola and Pinder 1982; Philip and Kitanidis
1989; Pardo-Igzquiza and Chica-Olmo 2004). In situ measurement of VHG is performed using
piezometer nests, which contain multiple piezometers in close proximity at different depths.
However, installing piezometers is expensive and time-consuming, and so the number of
piezometers at a particular site is usually limited, and piezometer nests are less frequently made.
For these reasons, VHG maps are often produced, but are typically limited to a relatively local area,
such as fluvial channels (for quantifying surface/groundwater interaction) and polluted sites (for
monitoring contaminant plume). On a wider regional scale, VHG mapping is rarely performed, and
delineation of regional groundwater flow systems is often just a 2D horizontal map with schematic
29
30
measurement errors (e.g., Lamotte and Delay 1997). It is important for kriging to take into account
measurement errors such as uncertainty in spatial variability (Saito and Goovaerts 2002). This
poses a conflict in the analysis of conventional well data. Available data, as much as possible, are
collected to compensate for reliable measurements of limited number, while such a dataset includes
larger variance due to wide-ranging measurement errors. For these reasons, an appropriate filtering
process is required to strike a balance between the volume and uncertainty of the dataset.
Here, the Toyohira River alluvial fan, Sapporo, Japan, is used as a case study on VHG mapping.
Over 1,000 water levels are compiled, and then a portion of the data is extracted by using long-term
water level analysis in observation wells. Water levels in individual wells are also grouped by well
depth into shallow wells and deep wells. A GTE map is first produced using the shallow well data.
Individual VHG values in the deep wells are calculated from their water levels, prior estimated
GTE, and representative values of screen depth elevation. There are two problems in this process.
(1) VHG is, in typical practice, calculated to determine the screen elevation between the top and
bottom elevations, which are often separated by several tens of meters or more. In this study, three
cases, namely, the top, middle, and bottom elevations are used, and then the resulting maps are
compared. (2) Regional mapping of GTE and VHG includes a non-stationary problem. In this
study, the deterministic drift in GTE is formulated as a linear relation with the ground surface
elevation. Moving neighborhood kriging is also used in VHG mapping. Cross-validation is
conducted to evaluate the validity in the GTE and VHG maps. The actual groundwater flow system
in the fan is examined from the resulting GTE and VHG maps.
2.2
2.2.1
Numerous water wells and boreholes have been constructed in and around the Toyohira River
alluvial fan. The target region including the fan is defined in the range of the X coordinate
(north-south) as 101,000 to 111,000 m and the Y coordinate (east-west) as 77,000 to 68,000 m
on the Japanese coordinate system (the Japanese Geodetic Datum 2000: JGD2000). The extent
loosely corresponded to latitude from 430513 to 425952N and longitude from 1411815 to
1412458E. In this study, a total of 1,392 water level data are firstly compiled (Figure 2.1). A
31
portion of these data are publicly available in publications of the Geological Survey of Hokkaido
(Yamaguchi et al. 1965; Obara 1969) and a digital shape file of the Hokkaido region (National and
Regional Policy Bureau 2007). Other data are from private or unpublished reports. Since the 1970s,
groundwater level and quality in the fan have been monitored by the Hokkaido Regional
Development Bureau and the Geological Survey of Hokkaido. The observation wells are shown as
solid squares in (Figure 2.1, and details about the wells are listed in Table 2.1.
32
Figure 2.1 Location maps and topographic features of the study area: the Toyohira
River alluvial fan. Small points are the wells and boreholes that provided
water level data before filtering (n = 1,392). Solid squares are long-term
observation wells, details of which are listed in Table 1. Bold black line
shows the position of the cross section in Figure 2. Gray dashed lines are
subway lines constructed in the city. Figure 1c also shows ground surface
contours with 1 m interval from 5 to 80 m asl, and short lines along the
river indicate KP which denote the distance (in kilometers) along the river
channel upstream from the confluence with the Ishikari River.
33
Table 2.1 Long term observation wells in the Toyohira River alluvial fan
b
Well
OW1D
OW1S
OW2
OW3D
OW3S
OW4D
OW4S
OW5
OW6
OW7
OW8
OW9
OW10
OW11
OW12
OW13
OW14
OW15D
OW15S
OW16D
OW16S
OW17D
OW17S
OW18
OW19
OW20D
OW20S
OW21
OW22
OW23
Location
Londitude Latitude
04' 39" 21' 41"
04' 39" 21' 41"
04' 32" 23' 30"
04' 15" 21' 11"
04' 15" 21' 11"
04' 04" 23' 26"
04' 04" 23' 26"
03' 46" 22' 00"
03' 44" 20' 34"
03' 40" 20' 45"
03' 37" 19' 37"
03' 31" 22' 49"
03' 20" 20' 53"
02' 47" 19' 51"
02' 45" 20' 43"
02' 41" 21' 22"
02' 24" 21' 47"
02' 23" 22' 34"
02' 23" 22' 34"
02' 23" 21' 26"
02' 23" 21' 26"
02' 13" 21' 52"
02' 13" 21' 52"
02' 07" 20' 08"
02' 05" 21' 15"
01' 46" 21' 15"
01' 46" 21' 15"
01' 33" 21' 59"
00' 57" 21' 02"
00' 55" 21' 04"
GH
13.54
13.78
11.11
15.70
15.70
14.01
13.97
18.05
18.12
18.96
17.33
15.87
23.47
25.04
28.14
29.74
34.22
35.11
35.13
31.30
31.30
39.36
39.37
34.61
35.20
39.00
39.00
53.41
54.01
53.38
Measurement period
Start date End date
1-Jan-79 31-Dec-08
20-Jun-73 31-Dec-09
20-Jun-73 9-Mar-11
19-Mar-11 17-Aug-12
19-Mar-11 17-Aug-12
10-Mar-06 31-Dec-08
10-Mar-06 31-Dec-08
9-Jun-73 31-Dec-09
3-Dec-93 31-Dec-09
9-Dec-80 23-Mar-93
3-Jan-77 31-Dec-08
20-Jun-73 31-Dec-09
23-Oct-73 11-May-02
11-Sep-80 31-Dec-09
25-Oct-79 10-Mar-11
1-Jan-91 31-Mar-12
17-Dec-80 31-Dec-09
1-Jan-79 31-Dec-08
1-Nov-73 10-Mar-11
9-Jun-10 17-Aug-12
9-Jun-10 17-Aug-12
10-Mar-06 31-Dec-08
1-Jan-07 31-Dec-09
11-Sep-80 31-Dec-09
25-Sep-10 17-Aug-12
10-Aug-10 17-Aug-12
9-Jun-10 17-Aug-12
20-Jun-73 10-Mar-11
6-Jun-73 10-Mar-11
10-Mar-06 31-Dec-09
Screen depth
c
Class
(m)
31.5 37.0
10.0 18.0
11.0 16.5
43.0 44.0
10.0 15.0
44.0 48.0
10.0 14.0
11.5 17.0
27.0 32.5
27.0 32.5
54.5 60.0
4.5 13.5
14.0 19.5
9.5 15.0
8.0 19.0
17.2 30.2
16.5 22.0
28.5 34.0
14.5 20.0
62.0 64.0
2.5 7.5
22.0 30.0
15.0 19.0
9.5 15.0
4.0 5.0
99.0 100.0
2.5 7.5
8.0 19.0
7.7 13.2
25.0 26.5
D
S
S
D
S
D
S
S
D
D
D
S
S
S
S
D
D
D
S
D
S
D
S
S
S
D
S
S
S
D
Original names of wells in Japanese are omitted, and numberings are given from the north to the south.
D and S, respectively, denote deep and shallow wells in close proximity (i.e., piezometer nests).
b
Location is common at 43N or 141E, and GH denotes ground height (m asl). cD and S denote deep
and shallow wells, respectively.
34
Furthermore, in 2010 the authors observed water levels at some of the individual wells and in a
longitudinal distribution of river stages to obtain more valid data.
Figure 2.2 Photos of water level measurement in observation wells; a) OW20S, b) OW6
The records for each well or borehole contained all or some of the following items: location,
total depth, installed screen length(s) and position(s), geology, pumping and non-pumping (static)
water levels, and pumping or logging test results. Figure 2.3 shows that well depths and screen
lengths are variously distributed: most well depths range from 30 to 50 m, and the screen lengths
are almost evenly distributed until 50 m.
35
30
30
Depth (m)
a
60
90
120
20
Frequency
40
60
60
90
120
20
Frequency
40
60
Figure 2.3 Histograms of well depths (a) and screen lengths (b). Screen length
is the distance from the top to bottom of screens. The wells (n = 203)
are those used for analysis; the depths are over 20 m, and the water
levels are measured since 1988
Over the past 50 years, static water levels have been measured at various times, but often only
during initial construction. In most records, dates of observations are given in year only, not month
or date. The hydraulic head mapping firstly required the extraction of data representative of the
target period, which, in this study, is the present. The filtering process is based on the long-term
water level variations at the 30 observation wells listed in Table 2.1. From 1969 to 1994, three
subway lines were in turn constructed in Sapporo, and during this period heavy pumping was
repeatedly conducted for dry work. The annual mean variations in the observation wells are
evaluated to detect the dewatering period. The differences between daily means in each
observation wells and total mean after the period are also statistically represented as box plots to
assess daily fluctuations. The annual trends in observation wells are evaluated by using
nonparametric techniques: the Mann-Kendall test (Mann 1945; Kendall 1975) and Sens slope
estimation (Sen 1968). The traditional techniques have been widely used to test for randomness
against trends (e.g., Wen and Chen 2006; Soderberg and Hennet 2007). In this study, the trend
analysis is archived using a free Excel macro (Salmi et al. 2002).
The extracted water levels are divided into two categories: shallow wells of up to 20 m deep and
deep wells of greater than 20 m deep. Groundwater levels in the shallow wells are considered to be
nearly equal to the groundwater table at the time and location, because the water tables are
36
observed initially at 20 m depth during the drilling of wells and boreholes in the fan. On the other
hand, groundwater levels in the deep wells might be affected by VHG at the location.
The data set is input into a geographical information system (GIS). GIS makes it feasible to
interpolate environmental variables through built-in applications based on deterministic or
geostatistical techniques (Burrough and McDonnell 1998; Johnson 2009). In this study, ArcGIS
v.10 (ESRI, Inc.) is used for analysis and mapping.
2.2.2
Kriging interpolation
Kriging gives the BLUE and its variance at any given point by using neighbor measurements and
a variogram model. Maps of target variables, GTE and VHG in this study, are generated by kriging
interpolation on a regular grid cell. The accuracy of the maps depends on the number and distances
of reliable measurements, and on the variogram model. The experimental variogram, which
characterizes the spatial variability of the data, is defined as half the average squared difference
between two attributed values (Deutsch and Journel 1998). In this study, variogram modeling (i.e.,
production of the experimental variogram and parameterization of the theoretical variogram) is
performed by using the software Surfer 10 (Golden Software).
Ordinary kriging and simple kriging are fundamental techniques in geostatistics. The difference
between them is small, but ordinary kriging is more useful in practice because ordinary kriging
filters the unknown mean required for simple kriging by adding the constraint that the kriging
weights sum to one. Ordinary kriging assumes stationary conditions where the mean of the
unknown is constant. If the unknown is not stationary but locally stationary in a certain
neighborhood, ordinary kriging is also applied within the local area (de Marsily 1986; Goovaerts
1997; Deutsch and Journel 1998; Wackernagel 2003). Moving neighborhood kriging is conducted
by setting a search radius less than the distance in which the stationary assumption is satisfied. The
groundwater table forms a subdued replica of the topography, and GTE may not even be local
stationary on a wider regional scale. For example, cokriging has been applied for estimating GTE
(Hoeksema et al. 1989; Chung and Rogers 2012). In other approaches, a trend component (drift) is
formulated as a function of an auxiliary variable. By definition, regionalized variable Z(x) is
conceptualized as a sum of a grobal trend component (drift) m(x), a spatial auto-correlation
37
(2.1)
Universal kriging and kriging with external drift have been performed (Aboufirassi and Mario
1983; Desbarats et al. 2002), in which kriging weights and drift parameters are simultaneously
solved in the kriging system. Alternatively, the drift and residuals are fitted separately and then
summed; this approach is applied in various techniques such as residual kriging, kriging combined
with regression, and regression kriging. The applicability also has been indicated through
comparison studies (e.g., Knotters et al. 1995; Hengl 2007; Nikroo et al. 2010). The formulas and
parameters of drift are firstly determined by ordinary least squares (OLS). Statistically, unbiased
parameters of drift are also determined by the iterative process of generalized least squares (GLS)
(Cressie 1993; Hengl 2007). GLS is a sophisticated but laborious process. The first OLS drift
should be satisfactory but slightly inferior (Kitanidis 1993; Minasny and McBratney 2007). In this
study, a simple linear relation is established between the GTE at each shallow well and its ground
surface elevation. The relation is defined as the drift of GTE. The GTE value is thus represented as
GTEi = m i + i = aGSEi + b + i,
(2.2)
where GTEi is the groundwater table elevation estimated at a shallow well, i is the index of the
shallow well, mi is the deterministic drift derived from the topography, i is random spatially
correlated residual, and GSEi is the ground surface elevation at the shallow well. Elevation in this
paper is given in meters above mean sea level. If GSEi in many wells is not available, GSEi is
estimated from a high-resolution digital elevation map (DEM) by using the GIS application. The
spatial variability of residuals, i is estimated as a variogram model, and residual map is produced
by ordinary kriging assuming a stationary condition. In this study, the grid size used is 100 m by
100 m. The size is set based on the accuracy of well location data. The drift map is prepared by
combining DEM and Equation (1). The GTE map is finally obtained by adding the residual map to
the drift map. The raster calculation in GIS is used for this process. The groundwater table depth
(GTD) is also mapped by subtracting the GTE map from the DEM.
Following GTE mapping, the VHG value at each deep well is calculated as
VHGj = (WLEj GTEj) / (GTEj SDEj),
(2.3)
where VHGj is the individual VHG value in a deep well and j is the index of the deep well. A
38
positive (negative) value indicates upward (downward) groundwater flow. WLEj is the water level
elevation, GTEj is the GTE extracted from the prior GTE map through the GIS application. SDEj is
the representative value of screen depth elevation between the top and bottom screen depths. SDEj
is considered to differ among the deep wells, and its determination is problematic. In this study,
three cases, namely, the top, middle, and bottom elevations of screens, are used for individual VHG
calculation (Figure 2.4). The VHG values in the different cases are respectively mapped by
ordinary kriging interpolation. The experimental variograms show some probability of
nonstationary in VHG, so moving neighborhood kriging is applied.
Deep
well
Shallow
well
G.L.
interpolation
VHGj
Depth=20m
SDEj
top
SDEj
middle
WLEj
GTEj
GTEi
SDEj
bottom
0
(m asl)
2.2.3
Cross-validation
Cross-validation is used to verify the resulting GTE and VHG maps. For the GTE map,
cross-validation is performed for residuals at all of the shallow wells, and then the residuals are
translated to GTE estimates by summing the topographic drift. The resulting GTE estimates are
compared with the measurements, GTEi. For the VHG maps, cross-validation is conducted at all of
39
the deep wells in each case of different SDEj values. The mean error (ME), root mean square error
(RMSE), and mean square standard error (MSSE) are used for error estimation:
ME
1 n
z ( xi ) z ( x i )
n 1
RMSE
MSSE
(2.4)
1 n
z ( xi ) z ( x i )
n 1
1 n z ( xi ) z ( x i )
n 1
i2
(2.5)
(2.6)
where n is the number of wells (shallow wells for GTE mapping, or deep wells for VHG
mapping), z(xi) is the measured GTE or VHG value, z ( xi ) is the estimated GTE or VHG using
the n1 other data, and i2 is the kriging standard deviation, which is the prediction error at
location xi. The selected variogram model and its optimized parameters are adequate when the ME
and RMSE values are close to zero, and the MSSE value is close to one with a tolerance range
[ 1 3 2 n , 1 3 2 n ] (Chils and Delfiner 1999).
2.3
2.3.1
The annual mean water variations at the observation wells are shown in Figure 2.5a. The
groundwater level variations show drawdowns in two typical seasons, i.e. winter and summer, the
former related to low recharge due to snow covering, and the latter due to (Fukami 2010).
Significant drops in water level are seen in most observation wells during 19851987, which are
years when subway construction occurred. The dewatering is not obvious in the apex at OW18 and
OW22U, which are located far from the subway lines. Water levels almost completely recover in a
few wells starting from 1988, but in most wells the levels remain low.
Figure 2.5b shows box plots of the differences of daily means from the total mean in each
observation well since 1988. The box plots statistically show daily fluctuations after the subway
construction, and water levels from that year are within a range of a few meters except in special
cases, such as heavy precipitation.
40
Figure 2.5 (a) Long-term variations in annual mean water levels and (b) box
plots of groundwater levels in the observation wells. Variations are
produced in the observation wells from hourly data covering at least
3 years. Box plots indicate differences of daily mean levels from the
total mean level in each observation well since 1988. Boxes from the
left to the right correspond to the numbering of the observation wells
(Table 2.1). The bars represent, from above, values of upper 90%,
upper quartile, mean, lower quartile, and lower 90%
41
The results of the Mann-Kendall test and the Sens slope estimates since 1988 are summarized
inTable 2.2. Positive trends are seen in an area of the distal part, and negative trends in an area of
the apex.
Positive trends likely indicate a change of water intake (e.g., a decrease in pumping rate) around
the distal part. Conversely, negative trends indicate
decreasing total water storage in the fan. In future work, we
will discuss the fans groundwater vulnerability and its
Wella
Slopeb
Significancec
mostly less than 0.05, and the largest is no more than 0.07.
OW1D
0.039 *
OW1S
0.036 *
OW2
0.034 ***
OW5
0.022 NS
OW6
0.032 NS
OW8
0.053 *
OW9
0.010 NS
OW10
0.001 NS
OW11
0.014 NS
OW12
0.014 NS
OW13
0.011 NS
OW14
0.060 **
OW15D
0.032 NS
OW15S
0.050 *
OW18
0.049 **
OW21
0.066 **
OW22
0.044 ***
42
2.3.2
3 h 1 h 3
2 a 2 a
h B CB
Here, B is the nugget effect, C is the partial sill, a is the range, and h is the separation distance.
Anisotropy of the spatial variability is a main consideration in hydraulic head mapping (Kitanidis
1993). In this study, isotropy in residuals is assumed because the topographic drift includes
anisotropy, and cross-validation results are satisfactory. The fitted model and optimal parameters
are shown in Figure 2.6.
Figure 2.7shows the cross-validation results. The ME and MSSE values are, in large part, close
to zero and one, respectively, indicating the suitability of the mapping. The RMSE value means
that the prediction errors are within several meters. Such estimation errors of this level are
regrettably inevitable under the uncertainty of conventional well data, owing to daily fluctuation
and long-term trends. However, the errors are insignificant for discussing the GTE map on a 101 m
order scale.
43
Figure 2.8 shows the estimated maps of GTE and GTD. Variance maps are omitted here, but the
above cross-validation shows that uncertainty is within acceptable limits. The steep land-surface
slope of the fan contributes to the substantial driving force from the apex to the distal part. The
GTE inclines northward between about 560 m asl. within a distance of only about 5 km. A GTE
mound and shallow GTD zone also appear along the river at KP 1419, indicating the interaction
between surface water and groundwater through the riverbed. The radius of the GTE mound and
shallow GTD zone is no more than 1 km from the river channel. The groundwater table is deep
(GTD 6 m) except around the river. Over 80 years ago, the groundwater table was shallower in
the Holocene fan, and the GTD was less than 05 m (Fukutomi 1928). In contrast to previously
published water table contours before subway construction (Yamaguchi 1983), the dewatering of
the fan exceeds 5 m from the middle to distal parts. In the southeast part of the Pleistocene fan, the
groundwater table is especially deep (GTD > 10 m), indicating an incorporation of groundwater
recharge by highland and dewatering by subway lines, as also indicated in the VHG map discussed
below.
44
Figure 2.8 Maps of (a) GTE and (b) GTD in the Toyohira River alluvial fan. Scattered black circles map shallow wells, boreholes, and observation
points for river stages. The black solid line denotes the topographic area of the alluvial fan, including the broken boundary line between the
Holocene fan and the Pleistocene fan. Short bars along the river denote transects with KP values. The gray lines indicate subway lines
through the fan. The broken sky blue lines are water table contours with 5 m intervals (m asl) in April 1960 before subway lines were
constructed (Yamaguchi 1983)
45
2.3.3
VHG map
Figure 2.9 Omnidirectional variogram and fitted spherical model for VHGj in
three cases: (a) top, (b) middle, and (c) bottom screen depth
elevations used for VHG calculation
46
Table 2.3 summarizes the results of cross-validating the deep wells in the three cases. The ME
values indicate an absence of bias in all the cases (Table 3a). The RMSE values are slightly
different among the cases, decreasing with greater screen depth. On the other hand, the variation in
MSSE values differs among the cases, and the value in the case of the bottom screen depths is only
within the tolerance interval [0.70, 1.30] (n = 203). The cross-validation results are also compared
in terms of error with respect to reliable VHG data (Table 2.3 b). In the distal part with small VHG
values (OW1 to 4), kriging estimates and variances insignificantly differ among the cases. In the
middle to the apex with negative VHG values (OW15 to 20), differences in the estimates are
relatively clear. Particularly at OW15 and OW16, the prediction errors are unacceptable in the case
of the top screen depths, but are relative small in the case of the bottom screen depths.
The comparison of cross-validations among the cases indicates that VHG is not constant and
hydraulic heads change inconsistently in the vertical direction, especially in the middle to apex of
the fan. As a result, the GTE map of only the bottom screen depths holds an acceptable accuracy
for such a 3D system. The approach for VHG mapping is strongly affected by which screen depth
is used as a representative position of water level, and the deepest value (i.e., the bottom of screen)
can be expected to reduce uncertainty because of the large denominator for individual VHG
calculations.
47
Table 2.3 Cross validation results for VHG mapping: (a) error estimators and (b)
Comparison with reliable VHG measurements at piezometer nests
a
Case
Error estimatorsb
ME
RMSE
MSSE
Top
0.00
0.19
2.67
Middle
0.00
0.13
1.55
Bottom
0.01
0.09
1.18
Cross-validation is conducted in three cases, the top, middle, and bottom screen depths for VHG
calculation. bME, RMSE, and MSSE are, respectively, mean error, root mean square error, and
mean square standard error (dimensionless).
b
Measurementsb
Cross-validation resultsc
Top
Wella
WLE
GTE
VHG
Est
Middle
Var
Est
Var
Bottom
Est
Var
OW1D&S
7.95
8.13
-0.01
0.04
0.02
0.02
0.01
0.01
0.01
OW3D&S
9.19
9.37
0.00
0.06
0.02
0.03
0.01
0.02
0.01
OW4D&S
6.14
5.81
0.01
0.07
0.02
0.06
0.02
0.04
0.01
OW15D&S
18.69
19.56
0.06
0.38
0.02
0.23
0.01
0.20
0.01
OW16D&S
20.51
28.17
0.13
0.32
0.01
0.21
0.01
0.17
0.01
OW17D&S
24.05
27.73
0.26
0.28
0.01
0.28
0.01
0.22
0.01
OW20D&S
17.32
36.78
0.20
0.26
0.02
0.19
0.01
0.14
0.01
Deep (D) and shallow (S) wells are installed in close proximity as piezometer nests. bWLE is water
level (m asl) in the deep well, and GTE is groundwater table elevation (m asl) in the shallow well.
GTE and WLE are mean values of all hourly data from 1988. Individual VHG values
(dimensionless) are calculated by Equation (2) using the middle screen depths in the deep wells.
(2). cEst and Var indicate kriging estimates and variance (dimensionless), respectively.
48
49
basement, in other words the sudden increase in the thickness of aquifer, probably induces an
accumulation of downward groundwater flow at the head. This highlights the potential
applicability of the VHG map in obtaining information about hydrogeologic structure at
considerable depth, which is usually accomplished by other geophysical methods such as seismic
investigations.
50
Figure 2.10 Maps of (a) VHG and (b) VHG variance in the Toyohira River alluvial fan. The elevations of top screens in the deep wells (circles) are
used for individual VHG values. The cross-validation (Table 2.3) showed the maps are statistically inferior than Figure 2.12.
51
Figure 2.11 Maps of (a) VHG and (b) VHG variance in the Toyohira River alluvial fan. The elevations of middle screens in the deep wells (circles) are
used for individual VHG values. The cross-validation (Table 2.3) showed the maps are statistically inferior than Figure 2.12
52
Figure 2.12 Maps of (a) VHG and (b) VHG variance in the Toyohira River alluvial fan. The elevations of bottom screens in the deep wells (circles) are
used for individual VHG values. The cross-validation (Table 2.3) shows the maps are more valid than Figure 2.10 and Figure 2.11
53
2.4
Mapping of vertical hydraulic gradient has rarely been attempted on the regional scale because
of the lack of piezometer nests and the uncertainty in conventional well data. This study
generalized the concept of the traditional plot method, and performed mapping of GTE and VHG
by kriging interpolation of conventional well data. The case study site was the Toyohira River
alluvial fan, Sapporo, Japan. A large number of water levels were observed by appropriate filtering.
The long-term variations in water level and box plots of its fluctuations in 30 observation wells
were produced, and nonparametric trend analysis was performed. Annual mean variation indicated
that construction of subway lines induced significant dewatering before 1988. Thereafter, the daily
fluctuations and the long-term variations ranged within a few meters. Water well data from 1988
were divided into those for shallow wells of up to 20 m deep, and deep wells of over 20 m deep.
First, the GTE was mapped by adding the topographic drift to the residuals, which were
interpolated by ordinary kriging. The drift was formulated as a linear relation between water levels
in the shallow wells and the ground surface elevation by ordinary least squares. Next, individual
VHG values in the deep wells were calculated by using its top, middle, or bottom elevations of
screen depth, respectively. VHG and its variance maps were also generated by using ordinary
kriging with a search radius fixed (moving neighborhood kriging). The cross-validation indicated
the validity of the GTE map and the VHG map of the bottom screen depths. The other VHG maps
of the top and middle screen depths could not be satisfactorily cross-validated. The VHG map, over
a wide area, visually reveals the magnitude and extent of recharge and discharge in the fan. The
downward flows are predominantly distributed over the fan due to the basement depression at the
head, the seepage loss from the river in the middle, and the artificial dewatering in the distal part.
The proposed incorporation of VHG and GTE maps provides a quasi-3D representation, under
the assumption that VHG at any location is invariable from the groundwater table to the bottom
of the aquifer; The assumption is not realistic. Several recent studies have currently challenged
three-dimensional mapping (Tonkin 2002; Rivest 2008). A 3D-representation of groundwater
head remains in future work.
54
Introduction
Groundwater recharge is the entry into the saturated zone across the water table surface, and
information about both recharge area and rate is critical to determine water balances and capacities
of aquifers (Bower 1978). Groundwater recharge is also important as upper boundary conditions in
numerical models of groundwater flow. For many regional groundwater studies, however,
insufficient attention is paid to recharge estimation (de Vries and Simmers 2002; Ruston 2003). This
is because groundwater recharge generally varies in space and time, depending on both systematic
and random fashions. Hence, estimation of recharge is most problematic in groundwater modeling
study as geologic heterogeneity.
A variety of techniques for recharge estimation are developed on different time- and space- scales
(e.g. Gee and Hillel 1988; Scanlon et al. 2002; Healy 2010). Most methods are fundamentally based
on the water-budget equation, such that recharge is calculated as a residual component in
precipitation from other meteorological, hydrological, and hydrogeological components.
Considering only measurement error, variance of recharge component is represented as a sum of
variances of other components. If recharge is small relative to other components, as usual there are,
uncertainty of recharge component can easily exceed 100% of the true value (Gee and Hillel 1988).
Thus, the validity of recharge estimation depends on accuracy of other components. However, all of
various components that consist of the water-budget equation are rarely obtainable with sufficient
accuracy level, especially on regional scale.
The water budget method is applicable under specific conditions that as much components as
possible are negligible. Intermittent focused recharge from streams or rivers is often the dominant
recharge mechanism in arid- and semi-arid regions (Healy 2010). Here, focused recharge is more
localized recharge from specific bodies of water to underlying aquifer, while diffuse recharge is
more regional recharge reffered from precipitation and snow melting. The rate of focused recharge is
generally higher as either slope or permeability of the river-bed increases, such as in the rivers
flowing on alluvial coarse aquifers. Also in such regions, river managements during low flow
periods are frequently focused on how minimum discharge under the seepage loss is needed to
55
rivers of low discharge. Typically, discharge measurements in gravel-bed rivers include more
uncertainty due to unsteady turbulent flow and roughness of boundary (Wiberg and Smith 1991).
Accordingly, the synoptic survey has been considered to be less applicable in the gravel-bed rivers.
This dissertation advances the synoptic survey method in the context of reduction of uncertainty
in the conventional flow measurement. The equation of flow uncertainty is indicated in ISO(2007),
and the equation consists of various factors such as point velocity, water depth, number of verticals,
and so forth. This research typically focuses two factors; point velocity and number of vertical. The
velocimeter used in this thesis is a handheld acoustic Doppler velocimeter (ADV), which yields
much more accurate velocities even in shallow water depth conditions. Next, proper arrangement of
verticals is also designed in each transect as to minimize the uncertainty. The improved synoptic
survey is applied in the Toyohira River, such that the longitudinal discharge and the leakage are
quantitatively determined. This investigation also aims to reveal a relation between the river leakage
56
and the discharge in the Toyohira River. The relation has been reported during high discharge and rapid
variation of river stage such as flood flow; bank storage effect. The relation during low and steady flow
has been rarely discussed, but the previous synoptic surveys in the Toyohira River indicated the
relation.
3.2
3.2.1
The water budget equation for a reach of stream can be written as (Healy 2010):
Qswout = Qswin + P ETsw Ssw + Roff + Qtrib + Qinter Qseep
(3.1)
where Qswout is the streamflow (discharge) at the downstream end of the reach, Qswin is the
streamflow (discharge) at the upstream end of the reach, P is the precipitation falling on the stream,
ETsw is the evaporation from the stream, Ssw is the change in stream strage, Roff is the surface
runoff to the stream, Qtrib is the flow from tributaries within the reach, Qinter is the interflow from the
unsaturated zone, Qseep is the net gaining if negative or losing if positive as an exchange of water
between the stream and the subsurface.
Qseep is determined as a residual in Equation (3.1). The subsurface terms generally dominate
Equation (3.1), and the measurement times are often selected so that precipitation, runoff, and
interflow are approximately zero. ETsw and Ssw can be estimated independently, but for practical
applications on naturally flowing streams the magnitudes of these terms are generally quite small
relative to that of the surface-flow terms and often insignificant relative to measurement errors of
surface flow (Healy 2010). Hence, Equation (3.1) is written as a simplified form:
Qseep = Qswout Qtrib Qswin
(3.2)
The method is traditionally conducted through synoptic discharge survey, conducting using stream
gaging at several stream cross sections during a short period (Woessner 2000; Simonds and Sinclair
2002; Harte and Kiah 2009). The exchange of water, Qseep can be estimated as the difference
between inflow, Qswin and outflow, Qswout. All terms in Equation (3.2) are independently measured,
and then the variance of the exchange is defined as:
Var[Qseep] = Var[Qsw out] + Var[Q trib] + Var[Qsw in]
(3.3)
The above equation indicates that the resulting Qseep needs to be substantially larger in magnitude
57
than the sum of right terms, Var [Qsw out], Var [Q trib], Var [Qsw in], for quantitative discussion.
3.2.2
A method for measuring flow discharge in a river has been standardized by ISO (2007), and other
methods have also been established in each country such as Japan (Public Works Research Institute
2002) and USA (Whiting 2003; Turnipseed and Sauer, 2010). Most publications commonly suggest
that velocity measurement is elemental and firstly selected. The velocity method is the method that
calculates flow discharge by a sum of subsectional charges which are obtained as products of river
width, depth, and average velocity:
Q = qi =bi di vi
(3.4)
where Q is the total discharge, qi is the subsectional discharge in the i th segment; bi is the increment
width; di is the representative (mean) water depth; vi is the representative (mean) velocity. The standard
uncertainty (one standard deviation values, level of confidence approximately 68%) of discharge is
calculated by the following equation (ISO 2007) as:
m
u Q u m2 u s2
2
b d v u
i 1
2
b ,i
1
u d2 ,iu 2p ,i
ni
bi d i v i
i 1
2
u c ,i u e2,i
(3.5)
where u(Q) is the relative (percentage) combined standard uncertainty in discharge; um is the uncertainty
due to the limited number of verticals; us is the uncertainty due to variable responsiveness of the
current-meter (ucm), width measurement instrument (ubm) and depth sounding instrument (uds), us also
corresponds to a root of sum of square of these terms. ub,I, ud,I, uv,i are the relative (percentage) standard
uncertainty in the width, depth and mean velocity measured at i th vertical, up,i2 is the uncertainty in mean
velocity vi due to the limited number of depths at which velocity measurements are made at i th vertical,
ni is the numbers of depths in the i th vertical at which velocity measurements are made, uc,i is the
uncertainty in the velocity at a particular measuring point in i th vertical due to lack of repeatability of the
current-meter, ue,i is the uncertainty in point velocity at a particular depth in i th vertical due to velocity
fluctuations in the stream during the exposure time of the current-meter. This research aims to improve
the method of flow measurement to reduce as much uncertainty as possible under the flow
58
59
Figure 3.1 Current meters and Handheld ADV Diagram of the FlowTracker probe.
bi d i v i
i 1
discharges, qi=bi di vi are equal. This is known as the statement of the Cauchy- Sharwtz inequality.
Additionally if m is enough large, the third term is ignored. However, it is impossible to design verticals
60
to satisfy the ideal statement because detailed discharge distribution in gravel-bed river is never
predictable before measurement. This dissertation suggest that verticals of velocity in each transect
b d v , are no
m
i 1
more than 0.1 (relative percentage: 10%), and number of verticals, m is enough large (m >20). This
criterion is expected to reduce as much uncertainty of the third term as possible.
The verticals of water depth measurement are arranged at the same location with those of water
velocity (ISO 2007). Another way is that number of verticals for measuring water depth is double
that of verticals the number of measurement by adding the verticals of depth at the middle points
between the adjacent verticals (Public Works Research Institute 2002). In gravel-bed rivers, adjacent
water depths are often variable, and the profiles of water depth have large fluctuations as shown in
Figure 3.3. Measuring water depth are more effortless than velocity measurements, so this
dissertation conducts 5-point means depth measurements; each subsections are equally divided into
five points, next water depths at 5-point are measured in turn, and the averages of five measurements
are calculated as a representative water depth in the subsection (Figure 3.3).
The number of depths in the i th vertical at which velocity measurements made, ni, affects on the
term in the right side of Equation (3.3), u
p ,i
1
)(u 2 c,i u 2 e ,i ) . The uncertainty in the mean
ni
velocity at a vertical upi depends on the number of points velocity in each vertical, and varies ranged
61
between 0.5 % and 7.5 %. However, velocity measurement is time consuming, and the number of
point velocity in each vertical directly increases the total time of synoptic survey. Synoptic survey is
needed to perform during a short time (ideally at the same time) as discharges at upstream and
downstream points are invariable if no exchange along the reach. By arranging sufficient number of
verticals m, the right term including the uncertainty upi is expected to be neglected. For this reason, in
this dissertation, the number of point velocity ni are determined as the conventional method
(Turnipseed and Sauer 2010): vi = 0.5v0.2 + 0.5v0.8, if D > 0.45 m (two point method); Vi = v0.6, if D <=
0.45m (one point method); Vi = 0.25v0.2+ 0.5v0.6+ 0.25v0.8, if the difference between v0.2 and v0.8 in each
vertical exceeds to either half of v2i or that of v8i (three point method), where v0.2, v0.6, v0.8 are the point
velocity on each vertical at 0.2, 0.6, 0.8 of the depth below the surface.
3.2.3
Observation transects
The synoptic surveys were conducted during wadeable conditions of the Toyohira River at four
events: 1) 1415 September, 2010; 2) 2425 September, 2010; 3) 89 October, 2010; 4) 24 February,
2011.
10 observation transects (St. 1 to 10) were located from the confluence of the Yamahana River to the
gaging station, Kariki at the distal part. The observation transects are shown in Figure 3.4 . The
transect (St. 9) beneath the Azuma Bridge was added in the latter measurements on October and
February to identify change from losing to gaining. The transect (St. 6) was shifted 250m upstream (St.
6) to obtain more validity of the measurements on February. The measurements at transects (St. 3 and
St. 7) were also performed for quantifying losing rate on 23 December, 2010 and 31 May, 2011. The
water intake to the Sousei River is automatically recorded per hour by the Hokkaido Regional
Development Bureau, and the averaged value during each period of synoptic survey was used for
calculating Equation (3.2). The discharges in the tributary, the Shojin River was no more than 0.1 m3/s,
and the error in relatively small value are negligible. This research measured the tributary discharge
by the conventional method using the current profiler meter.
62
Figure 3.4 Map of Toyohira alluvial fan and gaging stations for synoptic surveys
63
3.2.4
64
65
3.3
3.3.1
The measurement results (date, number of verticals, total width, total area, mean depth, velocity
(mean and maximum)total dischargeISO uncertainty (relative percentage and discharge)) are
summarized in Table 3.1. The variations of river stages at the Kariki and Moiwa gaging stations at
any surveys were within 0.01 m, and the variations of groundwater level at the observation wells,
BW3 and BW5, were also within 0.010.03 m. The small fluctuation of river stage and groundwater
level indicated the validation of quasi-steady condition (Ssw 0). The number of verticals m in each
transects were 2240, which were double to four times than that in the conventional method (m= 10). As
a result, ISO uncertainty values obtained were only 23 %. The accuracy was obtained owing to
incorporation of the handheld-ADV (us= 1 %) and the detailed arrangement of verticals (um=12.5 %).
One example at the Kariki gaging station (St.10) at 3 February, 2011 is shown in Figure 3.7. The
profile of water depths indicated that the water depths were highly variable between 0 and 0.83 m,
and that the differences of water depth are often over 0.1 m even in the pairs of adjacent points.
Figure 3.7 shows that the profile of averaged water depths (open triangles) almost corresponds to the
profile of moving average of 0.1m interval measurements (red line). The agreement indicates that the
5-point mean depth measurement method is sufficiently applicable for measuring water depths in
the gravel-bed river. The verticals of velocity measurements were located with the interval of 1 m in
the left-side (distance of 2336 m), of 4 m in the middle (distance 3650 m), and of 0.5m in the
right-side (distance of 5061 m). The total number of verticals m of 40 was given. The measured point
velocities largely varied in the range from 0.18 m/s (minimum) in the middle of the channel to
maximum 1.05 m/s (maximum) in the right side. The contour map of velocity is seen as continuous
and smooth, indicating the velocity distribution was well obtained. The slight disturbance in 0.8 depth
velocities was recognized only in the right-side. The ratios of sub-sectional discharge qi to the total
discharge ranged between 0.4 and 5.3 %; mostly less than the target value (5%). Consequently, the total
discharge was 8.49 m3/s, and the uncertainty was only 1.9 %, which corresponded only to 0.16 m3/s.
66
Table 3.1 Measurement results by synoptic surveys using ADV in the Toyohira River
St.
Total
Velocity (m/s) Discharge ISO Uncertainty
Number of
Total
Mean
2
3
3
Verticals Width(m) Area(m ) Depth (m) Mean Max.
(m /s)
(%)
(m /s)
31
47
8.67
0.18
0.43
1.16
3.71
2.7
0.10
35
56
12.01
0.21
0.28
0.77
3.36
2.2
0.07
33
31
7.19
0.23
0.46
0.70
3.32
2.8
0.09
27
17.8
4.56
0.26
0.66
1.10
2.99
2.9
0.09
24
22.3
6.55
0.29
0.41
0.74
2.67
3.1
0.08
28
26
8.98
0.35
0.31
0.52
2.81
2.5
0.07
25
34.4
23.45
0.68
0.12
0.17
2.78
2.4
0.07
25
21.9
11.14
0.51
0.26
0.41
2.85
2.6
0.07
33
38.2
9.89
0.26
0.29
0.59
2.87
2.7
0.08
1
2
3
4
5
6
7
8
10
Date
15-Sep-10
15-Sep-10
14-Sep-10
15-Sep-10
14-Sep-10
14-Sep-10
14-Sep-10
14-Sep-10
15-Sep-10
1
2
3
4
5
6
7
8
10
24-Sep-10
24-Sep-10
25-Sep-10
25-Sep-10
25-Sep-10
25-Sep-10
24-Sep-10
24-Sep-10
24-Sep-10
37
37
34
27
23
27
26
25
33
47
56.2
32.3
18.2
22.2
25.3
34.4
22
38.1
8.80
10.92
7.91
4.11
6.47
9.20
23.14
11.02
9.76
0.19
0.19
0.25
0.23
0.29
0.36
0.67
0.50
0.26
0.34
0.28
0.39
0.71
0.43
0.30
0.11
0.23
0.27
1.08
0.72
0.66
1.11
0.73
0.50
0.17
0.37
0.55
3.03
3.08
3.10
2.92
2.79
2.80
2.62
2.51
2.66
2.5
2.5
2.5
2.9
3.1
2.6
2.4
2.6
2.7
0.08
0.08
0.08
0.08
0.09
0.07
0.06
0.07
0.07
1
2
3
4
5
6
7
8
9
10
8-Oct-10
9-Oct-10
9-Oct-10
9-Oct-10
9-Oct-10
9-Oct-10
8-Oct-10
8-Oct-10
9-Oct-10
8-Oct-10
38
40
32
26
31
30
25
25
22
34
47.4
56.2
32.8
18.7
24.5
26.6
34.5
22.5
34.6
38.3
11.43
13.61
10.34
5.82
8.23
10.85
25.38
12.96
21.91
11.65
0.24
0.24
0.32
0.31
0.34
0.41
0.74
0.58
0.63
0.30
0.46
0.35
0.48
0.76
0.49
0.39
0.16
0.31
0.18
0.39
1.16
0.82
0.78
1.35
0.76
0.67
0.24
0.54
0.26
0.66
5.29
4.75
4.99
4.44
4.06
4.22
4.08
3.99
4.04
4.49
2.3
2.0
2.5
2.6
2.6
2.3
2.4
2.5
2.7
2.4
0.12
0.09
0.12
0.12
0.11
0.10
0.10
0.10
0.11
0.11
2
3
7
10
24-Dec-10
23-Dec-10
23-Dec-10
24-Dec-10
33
36
27
32
56.80
34.30
52.10
43.70
23.60
17.33
30.86
19.11
0.42
0.51
0.59
0.44
0.61
0.79
0.42
0.74
1.18
1.26
0.65
1.35
14.38
13.61
12.90
14.07
2.2
1.9
2.3
2.1
0.32
0.26
0.30
0.30
1
2
3
5
6'
7
8
9
10
4-Feb-11
4-Feb-11
2-Feb-11
2-Feb-11
4-Feb-11
2-Feb-11
4-Feb-11
4-Feb-11
3-Feb-11
30
35
38
30
33
22
26
30
39
36.9
57.1
33.3
34.6
49.2
35.3
23.3
44.7
43.8
15.52
18.83
13.91
13.88
21.56
32.29
14.96
35.39
15.37
0.42
0.33
0.42
0.40
0.44
0.92
0.64
0.79
0.35
0.57
0.47
0.65
0.63
0.40
0.25
0.54
0.22
0.55
1.12
1.07
1.09
1.02
0.65
0.36
0.78
0.33
1.05
8.91
8.80
9.10
8.70
8.52
7.94
8.08
7.84
8.49
2.4
2.1
2.0
2.3
2.1
2.7
2.4
2.1
1.9
0.21
0.18
0.18
0.20
0.18
0.21
0.19
0.16
0.16
3
7
31-Mar-11
31-Mar-11
36
28
33.7
34.5
14.66
31.34
0.44
0.91
0.67
0.28
1.18
0.40
9.82
8.88
2.0
2.2
0.20
0.20
67
Figure 3.7 Example of measurement results of the Kariki Station (St.10) for 3
February 2011. Upper shows water depths; black line is 0.1 m interval:
red line is 1 m moving average: inverse open triangles are 5-point
mean, and velocity contours with squares obtained by ADV. Lower
shows each subsection discharge ratio to the total discharge (8.49
m3/s) at each vertical
3.3.2
The longitudinal discharge distributions from St. 1 to St. 10 are shown in Figure 3.8. All of the
profiles indicated the same trends although the magnitudes of discharges were different among the
surveys: 1) In the upper reach, the discharges slightly varied between St. 1 and St. 3, and the water
intake (0.30.4 m3/s) to the Sousei River are relatively negligible; 2) In the middle reach, the discharge
currently decreased from St. 3 to St. 7. The inflow of the Shoujin River (KP16.5about 0.1 m3/s) was
68
also ignored. The discharges between St. 7 and St. 8 were almost constant, and so the distinct losing
section of the River is determined as the reach from St. 3 to St. 7 (i.e. total length of about 1.5 km). The
distinct losing section was relatively short as compared with the total length of the river through the fan
(about 10 km). The river leakage occurred in the typical conditions to balance hydraulic gradient and
hydraulic conductivity in the riverbed.
; 3) In the lower reach, the discharges are nearly constant between St.7 to St. 9, indicating less
interaction between surface/groundwater. However, the discharges slightly increased between St. 9 to
St. 10, and the lowermost reach was identified as the gaining section, although this increase was
observed only twice.
69
Azuma Bridge
10
(outflow0.3-0.4 m /s)
(inflow0.1 m /s)
10
60
48
6'
8
River stage
(28-29 June 2010)
Distinct Losing Reach
17.0
6
1
3
10 (Kariki St.)
2
(Moiwa St.)
4
24-25 Sep. 2010
10
0
11
12
13
24
Stage (T.P.m)
36
Discharge (m /s)
15.5
14
15
KP(km)
2
12
16
17
18
Figure 3.8 Synoptic survey results; longitudinal discharge distributions of the Toyohira River. KP is the distance from the confluence of the Toyohira and
the Ishikari Rivers. Error bar represents the discharge uncertainty calculated using the ISO method. Solid blue line is the river stages in 2829
June 2010.
70
3.3.3
Exchange of water between the river and the subsurface was determined by Equation (3.2). In this
case, two tributaries, the Sousei River and the Shojin River, were considered for calculation as:
dQi to j = Qi Qj QSousei + QShojin
(3.6)
where dQi to j is the exchange of water between the upper and the lower transects. Positive values
indicated losing, and negative values indicated gaining. Qi and Qj were the discharges at the upper and
the lower transect, respectively. Index i and j corresponded to the station numbers as shown in Table 3.1.
QSousei was the water intake into the Sousei River, and the term is ignored if i>= 3 or j<= 2. QShojin was
the inflow discharge from the Shojin River, and the term was ignored if i>= 6 or j<= 5. The uncertainty
of dQ i to j was also calculated by using the upper and the lower discharges and their uncertainty, as
shown in Equation (3.3):
U i to j =
(U i Qi ) 2 (U j Q j ) 2
100
(3.7)
dQi to j
The calculation of exchange were performed between the three sections: the overall section
between the Moiwa (St. 2) and the Kariki (St. 10) gaging stations; the distinct losing section between
the Minami- 19 jyo Ohashi Bridge (St. 3) and the Minami- Ohashi Bridge (St. 7); the gaining section
between Azuma Bridge (St. 9) and the Kariki station (St. 10).
The estimated values and uncertainty are summarized in Table 3.2. Total exchanges through the fan,
dQ2to10, were commonly less than 0.15 m3/s, because that there are different reaches through the
overall section; losing section in the upper reach and gaining section in the lower reach. In such case
that exchanges of losing and gaining are close to each other, total exchange has difficulty to estimate
by using even this improved survey, and detailed measurements are needed by separating the long
reach into several sub-sections.
The exchange values of the distinct losing section, dQ3to7 were constantly positive, indicating the
losing. The scattered plots of the exchange values are shown in Figure 3.9. The leakage varied ranged
between 0.53 and 1.27 m3/s, and the variation was primarily attributable to the relative uncertainty.
Note that the two exchange values in the low flow conditions (<5 m3/s) were obviously smaller than
the other values during the relatively large flow conditions. This probably indicated the decrease in
permeability of the riverbed materials during the low flow conditions, and several factors were
71
considered in the literature, for example clogging on the riverbed (Katakai et al. 2006;
Goldschneider et al. 2007) and unsaturated zone between the surface water and the groundwater
table (Woessner 2000). However, the assumption was not valid only in the results, because the small
values become close to the other values if the relative uncertainty expands, for example, the
confidence interval changes from 68% interval to 95% interval. In the large discharge condition (>5
m3/s), a typical relation between the leakage and the discharge was not found in Figure 3.9. The
average values of dQ3to7 are 0.88 m3/s from the all values, and 1.04 m3/s from those except for the
two small values, respectively. The two average values are not significantly different, and it is valid
that the representative value of exchange in the distinct losing section, dQ3to7, was about 1m3/s, as
well as the previous surveys between the 1960 and the 1990. The similarity with the previous
surveys indicates that the river leakage might be almost invariable for over half a century. In addition,
the amount corresponds to about 80% of the total pumping rate (about 100,000m3/s) of groundwater
throughout Sapporo City. This means that the focused recharge plays an important role in
maintaining the groundwater reservoir, although the distinct losing section is only about one tenth of
the total reach through the fan.
The exchange of the third section, dQ9to10, obtained was only two values, but these values were
commonly estimated at about 0.5 m3/s. This means that the surface water was gaining between the
Azuma Bridge and the Kariki gaging station. The gaining forms springs in the riverbed, and the
springs are valuable for contributing to the spawning of salmon (Okamoto 2000).
72
1.5
1.0
0.5
Discharge Q3 (m /s)
10
15
Figure 3.9 Relation between river discharge and leakage. Q3 and dQ3to7 are
discharge at St.3 and leakage between St.3 and St.7. Error bars are
the estimated standard deviation of leakage.
73
Table 3.2 River leakage calculated using the stream water budget method with measurement results of the synoptic survey, positive value at water losing
and negative value at gaining. Q, U, dQ, and dQU respectively denote discharge, ISO uncertainty, rate of leakage, and uncertainty of leakage
Measurement Results
date
St.2
Moiwa
Q2
U2
St.3
Minami 19jyo Ohashi
Q3
U3
St.7
Minami
Ohashi
Q7
U7
Tributary Discharge
St.9
Azuma
Q9
U9
Q 10
Shojin
River
Pumping
to Sousei
River
(outflow)
(inflow)
U 10
Q Sousei
Q shoujin
St.10
Kariki
River Leakage
St.2 to St.10
3.36 0.07
3.32
0.09
2.78 0.07
2.87 0.08
0.42
0.08
0.15
0.11
0.62
0.11
3.08 0.08
3.10
0.08
2.62 0.06
2.66 0.07
0.38
0.05
0.09
0.11
0.53
0.10
8 - 9 Oct. 2010
4.75 0.10
4.99
0.12
4.08 0.11
4.49 0.11
0.33
0.07
0.00
0.14
0.98
0.17
13.61
0.34
12.90 0.30
0.29
0.13
0.84
0.45
14.07 0.30
0.29
0.13
0.15
0.43
8.49 0.16
0.28
0.11
0.14
0.25
1.27
0.28
0.37
0.11
1.05
0.28
23 Dec. 2010
4.04 0.11
8.80 0.18
9.10
0.18
7.94 0.21
9.82
0.20
8.88 0.20
7.84 0.16
-0.45
0.15
-0.65
0.23
unitm /s
74
3.3.4
75
Table 3.3 Comparison between results of synoptic surveys by ADV and conventional measurements by propeller at the Moiwa (St.2) and Kariki (St.10)
stations
St.
Moiwa
(St.2)
Date
15-Sep-10
24-Sep-10
24-Dec-10
Number of
Verticals
35
37
33
Kariki
(St.10)
15-Sep-10
24-Sep-10
8-Oct-10
24-Dec-10
3-Feb-11
33
33
34
32
39
0.290
0.273
0.386
0.736
0.553
2.87
2.66
4.49
14.07
8.49
2.7
2.7
2.3
2.1
1.9
12.52
12.69
11.29
20.06
16.12
0.25
0.35
0.44
0.62
0.64
3.16
3.30
4.95
15.25
9.48
8.9
8.8
8.9
7.4
8.5
1.10
1.24
1.10
1.08
1.12
76
77
Figure 3.10 Example of measurement results obtained using ADCP, at the Kariki
gaging station (St. 10) on 3 February, 2011. The result is the first
among the twice measurements at the same time of the ADV
measurements (Figure 3.10). Open squares are the measurement of
velocity normal to the tag line, and the color contour was made by
interpolating the point velocities. Lack of contour in the location of
41-50m is due to the shallow water depth for ADCP.
78
Azuma Bridge
(outflow0.3-0.4 m /s)
(inflow0.1 m /s)
1
2
5
(Kariki St.)
(Moiwa St.)
6'
Discharge (m /s)
10
10
9
the Higher
the Lower
6
11
12
13
14
15
KP(km)
16
17
18
Figure 3.11 Comparison of ADV and ADCP results. Open diamonds are the
results by ADV, and the error bar is the estimated standard deviation,
open triangles are the twice results by ADCP, and open squares are
the averages of the twice measurements.
79
3.4
Conclusions
A synoptic survey of the river flow was performed to quantify the focused recharge from the
Toyohira River. The complexity and variation of water depth and velocity distributions in the
gravel-bed river frequently caused unacceptable uncertainty due to the unsteady turbulent flow and
rough boundaries. For reducing the uncertainty as much as possible, this dissertation proposed the
two following improvements; 1) an application of a handheld acoustic doppler velocimeter (ADV) and
2) detailed arrangements of verticals measurements of flow and depth. Applied to the river, continuous
measurements were taken at 46 stations each day, yielding ISO uncertainty of discharge of
approximately 23%, that were reduced as much as possible. The synoptic survey showed its required
accuracy and practical availability. It revealed the longitudinal river discharge and leakage of the
Toyohira River. The distinct losing reach was only about 1.5 km between the Minami 19-jyo Ohashi
Bridge (St. 3) and the Minami Ohashi Bridge (St. 7). The mean leakage was estimated at about 1 m3/s.
The total budget between the upper and lower gaging stations, Moiwa and Kariki, was too little to
be observed because of a change in a river-leakage pattern from loss to gain around the Azuma Bridge.
Compared with the consistent flow conditions, the measurements using the conventional method with
the propeller-type current meter included higher uncertainty of no less than 79%. Thus, the
conventional method was not appropriate in quantifying the interaction between surface- and
groundwater in the river. Measurements obtained using the ADCP showed good agreements with the
measurements obtained using ADV in deep water conditions: depth greater than 0.3 m. In the low
discharge condition of the Toyohira River, however, the total discharge by the ADCP was lower in each
section because of the shallow water limitation of the instrument.
This proposed methodology contributes to yield more reliable estimates of exchange of water in
gravel-bed rivers than the conventional method. However, the accuracy is not enough, and the
exchange is even susceptible to the uncertainty. There are several points for addressing the
uncertainty; a repeat of flow measurements at the same flow conditions, statistical analysis,
subdivision of a long reach that is unknown whether gaining or losing. This method is applied only
in the low-flow and wadeable condition, but flood events are also of importance as a result of bank
storage and additional infiltration areas (Engeler 2011). The measurements of the exchange in the
high-stage, for example early spring with snowmelt, would be performed in a future work.
80
Introduction
Hydraulic conductivity (K) is most critical among various hydraulic parameters in groundwater
modeling because the parameter straight influences the groundwater flow fluxes according to Darcys
law. Even in a homogeneous aquifer, K is several orders of magnitude different among even
closely-separated points. The large variability of K in space depends on geologic heterogeneity
attributable to grain size, porosity, sorting, packing and so forth. Geologic heterogeneity also has
scale-dependence, and K varies on microscale (cores), mezoscale (facies), and macroscale (sequences).
Therefore, aquifer characterization in groundwater modeling studies almost coincides with a
description of spatial images of K in the analysis domain.
In situ measurements of K are directly obtained by various methods such as multiple pumping tests,
slug tests, flow meter tests and permeameter tests (Fetter 2001). However, the number of measurements
by in-situ methods was limited because of economical reasons. The measurements are also inevitably
affected on various degrees of uncertainty such as measurement error and scale-dependence.
Geostatistical approach was powerful to address the scarcity and uncertainty in such regionalized variable
(Dagan 1997; Eaton 2006; Lee et al. 2007). The approach assumes that the spatial variability is not purely
random, and that there is some kind of correlation in the spatial variability. The approach also requires
statistical assumptions, i.e. stationarity and ergodicity. Stationarity indicates that statistical properties such
as mean and variance are constant over the space, but the assumption is not apparent in geologic materials
at most scales (Anderson 1989).
Trending heterogeneity (Freeze and Cherry 1979) is also typically recognized in alluvial coarse
deposits with both enormous aggradation and steep land-slope. A downward decreasing trend of K is
widely known as a global trend in most fluvial terrains. Especially in alluvial fans, another humped
heterogeneity is generally observed (Cehrs 1979; Neton et al. 1994), where transmissivity reached at
highest values in the mid-fan area as a result of sorting trends. There are also coarsening/thickening or
fining/thinning sequences, which might indicate the correlation of K profiles (Neton et al. 1994).
The dependence of permeability on depth in consolidated and unconsolidated sediments is mainly due
to decreasing porosity because of compaction and other physical or chemical effects. Vertical
81
permeability variations in each aquifer are key information for quantifying groundwater flow and solute
transport. For example, the typical information aids in realizing geological heterogeneity in geostatistical
methods, which often require a fundamental assumption of stationarity (de Marsily 1986; Koltermann
and Gorelick 1996; Wackernagel 2003). Analytical and numerical solutions indicate the importance of
depth-dependent hydraulic conductivity K in groundwater systems of various scales (Saar and Manga
2004; Marion et al. 2008; Jiang et al. 2009; Cardenas and Jiang 2010; Zlotnik et al. 2011). A systematic
decreasing trend in either porous media or fractured media is can also be understood by using
semi-empirical models based on simplifications and well-established relations among permeability,
porosity, fracture aperture, and effective stress (Jiang et al. 2010).
In contrast to many theoretical approaches, semi-empirical approaches rarely give a unique depth
dependence of K at a particular site, especially in alluvial shallow aquifers, because geologic
heterogeneity prevents detection of a specific trend from a large variability in K. Another reason is a lack
of knowledge about the process of porosity structure development under various sedimentary
environments. In addition, the numbers of measurements and samples are limited in the majority of
practical cases. Vertical sequences in alluvial aquifers generally consist of various facies (e.g., Miall 1992),
for which there are a variety of physical and chemical effects on the permeability variations with depth.
Characterization has proved especially difficult in a complex alluvial fan system. The stratigraphic
sequences exhibit unique sedimentary textures on different scales due to various proportions of fluvial
flow, debris/mud flow, sheet flood, and sieve deposition (Einsele 2000). Coarsening- and fining-upward
trends are also typically seen in alluvial fans as a result of interactions of progradation, retrogradation, and
basin subsidence (Neton 1994). The vertical trend of grain size may positively correlate to the depth
dependence of K. However, grain size distributions alone are insufficient for determining the vertical
trend because K in the coarse sediment mixtures is a complex function of various geologic factors. An
additional problem is uncertainty in grain size analysis using gravel cores of small diameter less than the
maximum grain size. Groundwater flow or transport modeling in gravelly aquifers is thus often
conducted under the assumption that K is approximately invariant in the vertical direction.
The depth dependence of K in unconsolidated gravel deposits has rarely been discussed because
porosity reduction by compaction is of less importance in clean coarse sediments than in fine sediments,
and determined relations between porosity and K are seldom adhered to in natural coarse sediments
82
(Morin 2006; Kresic 2007). Previous studies on the depth dependence of K were performed mainly for
consolidated rocks because diagenetic and metamorphic affects occur only at depths on the kilometer
scale. The results of those studies were commonly that K exponentially or logarithmically decreased with
depth, and that decay exponents varied on the order of 1 10-2 to 1 10-4 m-1 at specific sites (Manning
and Ingebritsen 1999; Saar and Manga 2004; Ingebritsen et al. 2006; Jiang et al. 2009; Wang et al. 2009;
Luo et al. 2011). K in unconsolidated gravel deposits varies greatly even under a slight change in porosity.
The fractional packing model (Koltermann and Gorelick 1995; Kamann et al. 2007), which represents K
in sediment mixtures as a function of the porosity and volume fraction of each component, indicates that
K ranges over several orders of magnitude by these factors. Major (1997; 2000) performed triaxial
compression and permeability tests on various poorly sorted debris flow sediments, and reported that the
permeability of debris flow mixtures varies exponentially with porosity, and that changes in porosity of as
little as a few percent could cause greater than 10-fold changes in permeability. Matsumoto and
Yamaguchi (1991) found that in situ measurements of K in Holocene gravelly deposits decreased by
about one order of magnitude for test fills of less than 7 m in height. They also conducted laboratory tests
using undisturbed gravelly samples, and proposed that the relations between permeability and effective
stress reflected the connectivity and compaction of the open pores. As another example, Chen (2011)
investigated channel sediments in the Platte River, USA, and hypothesized that a decreasing trend in
vertical K with depth resulted from hyporheic processes that moved fine grained sediments from shallow
parts of the channel to deeper parts. All of those studies concluded that K is highly dependent on depth in
unconsolidated gravel deposits. However, the trends were usually shown only graphically, and more
quantitative characteristics about the trends (e.g., the decay exponents) were not typically determined.
The sedimentary or permeability correlation length in the vertical direction is within only a few meters
or even decimeters (Hess et al. 1992; Jussel et al. 1994; Rubin 2003; Falivene et al. 2007); thus, a number
of measurements are required to distinguish the vertical trend of geological heterogeneity. As an
alternative, indirect methods using empirical equations of grain size, porosity or other factors in samples,
give K values that approximately agree with measurements taken at the same depth (e.g., Vukovic and
Soro 1992; Cheong et al. 2008; Song et al. 2009; Vienken and Dietrich 2011). Discrepancies in indirect
methods are considered to be caused by the simplification of equations to only a few variables (or even a
single variable, for example, effective grain size diameter).
83
Undisturbed core sampling of gravel deposits is applied to investigate properties in the deep zone.
Undisturbed sampling of gravel deposits is challenging even today. In situ freezing of samples is
commonly employed, but the number of samples is restricted with this method due to its high economic
cost. In contrast, several nonfreezing sampling techniques have been developed (e.g., McElwee et al. 1991).
Unique to Japan, various tube samplers have been improved for high-quality sampling of gravel deposits.
Tanaka et al. (1990) indicated the similarity between measured physical and mechanical properties of
gravelly samples obtained by these improved tube samplers and by in situ sample freezing. The diameters
of undisturbed samplings are usually less than 1 m, and are not enough to evaluate the true
distributions of grain size. ASTM determines that grain size analysis needs the sample of diameter
larger than several times of maximum grain size. The gravel core of small diameter is also used to
determine sedimentary structures related to hydro-facies in gravel deposits. However, the small
undisturbed cores keep how fine sediments are packed between the gravel grains.
This dissertation suggested that when applying an indirect method for unconsolidated gravel deposits,
other variables are needed that reflect the packing in openings between the gravel grains. Consequently,
an additional matrix packing level index was proposed using relatively undisturbed gravel cores
obtained from the Toyohira River alluvial fan, Sapporo, Japan. A distinct difference is found in the
undisturbed cores obtained by the samplers in terms of the packing of fine sediments in pore spaces
between gravels. The packing in undisturbed cores is qualitatively categorized into four typical levels, and
the length fraction of each packing level was measured. The equivalent horizontal value of K per unit
depth was then formulated by using the grain size diameter and length fraction of each packing level.
Next, an additional grain size analysis is performed on past core samples, and profiles of the grain size
diameter are found for three sampling points, two in the mid-fan and one in the fan-toe. Secondly, the
previously determined relation is applied to convert the core property profiles into those of the estimated
conductivity. To eliminate errors during this process, the moving averages of log-conductivities are
calculated and the decay exponent of K with depth is estimated through a linear regression analysis.
Finally, a longitudinal cross section is then generated from the profiles such that the boundaries of the
gravelly aquifer structure are determined.
84
4.2
4.2.1
Figure 4.1 Schematic of improved double core-tube sampler used in this study
(ACE Shisui, Co., Ltd., Japan)
85
I: full
II : almost full
IV : very loose
K= 2.1510-3m/s
K= 4.3810-6m/s
III : loose
Figure 4.2 Relatively undisturbed cores, obtained at BW03 between 25 and 30 meter in
depth, showing examples of matrix packing level and slug test result
86
Figure 4.3 Location maps and geologic features of the study area; the Toyohira River
alluvial fan. Solid circles represent undisturbed sampling points; BW17;
dashed lines denote water surface contours with elevation values (in
meter above sea level) measured in June 2010; KP values denote the
distance (in km) along the river channel upward from the confluence of the
Ishikari River
87
(4.1)
where K is the radial or horizontal hydraulic conductivity, De is the effective radius of the well casing, D is
the test depth diameter, Kz is the vertical hydraulic conductivity, = K Kz is the conductivity ratio
representing the anisotropy, H is the test screen length and s1 (s2) is the drawdown at time t1 (t 2). Conversely,
if the water level variation is too rapid to be measured manually due to high permeability in the test section,
a pumping test is instead conducted in the borehole, and the pumping discharge rate and drawdown from
the static level in the borehole are measured under the pseudo-steady state. In this case, K is calculated as:
K Q 2sH log(H D 1 H D ),
2
(4.2)
where s is the drawdown under the pseudo-steady state and Q is the pumping discharge rate. In all tests, the
value of H was consistent at unit depth. The water temperatures during each test were measured by
transducers. The temperature was in the 512 C range, and was related to the infiltration near the river.
Figure 4.4 shows a relation between slug test results and test depths, indicating depth-dependence in
the gravel deposits.
88
Depth (m)
20
40
60
80
100
-5
-4
-3
Log10 K (m/s)
-2
Figure 4.4 Relation between previous results of slug test and mid-depths of the
test screens in the fan
89
Figure 4.5 Classification of gravel core according to matrix packing level. Open
granular squares in the column denote gravels, small dots denote fine
sediments in openings between gravel grains and black shadows
denote losses of fine sediments. Packing levels I and II (III and IV) at
unit depth are grouped as the high (low packing part)
90
Core sample
I: full
Feature
All pore spaces between gravel grain are
fully packed by fine sediments. No cracks
are identified througout the core.
III: loose
Figure 4.6 Box and whisker plots showing the vertical statistics of length fractions of
low packing parts a. L3 and b. L4. Summaries are given at 10 m intervals.
Solid squares above each box denote measurements at BW1-7, the leftand right-hand side bars of the boxes respectively denote the 25th and
75th percentiles, the bars within the boxes denote the medians, the bars at
the left and right ends of the whiskers respectively denote the minimum
and maximum values and open squares in the boxes denote the averages
91
4.2.2
The following relations between the 32 slug test results and core properties at the same depth are
established. (1) The K value of the high packing part (i.e., packing levels I and II) in the core is assumed to
be that of a conventional porous medium. The fundamental formula for K in porous media is generally
represented as a composite of the mediums fluid properties, porosity function or coefficient, and grain size
(e.g., Freeze and Cherry 1979; Todd and Mays 2005):
K=
g
Cd 2 ,
(4.3)
where is the density of the fluid; g is the acceleration due to gravity; is the kinematic coefficient of
viscosity; C is a dimensionless coefficient, which is dependent on the porosity, sorting, packing and other
factors; and d is the effective grain size diameter. d10 is commonly used for d (Fetter 2001); however,
Shepherd (1989) suggested that the mean diameter better represents the effective grain size diameter, and
the exponent ranges from 1.5 to 2 according to the sedimentary textural maturity. In the current study, the
obtained grain size distributions are noted as being different from those in nature because of sampling
limitations; namely, the sampler was of small diameter and the small sample volume were used for sieving.
Therefore, a more general exponential equation was applied to determine K in the high packing part:
K HP C HP d m , (4.4)
92
where KHP is the hydraulic conductivity of the high packing part, constant CHP is a dimensionless
proportionality coefficient and m is the exponent for effective grain size diameter d. Note that CHP
corresponds to the product of the fluid properties at a constant temperature of 25 C and the coefficient C in
Equation (4.3).
(2) The low packing part (i.e., packing levels III and IV) in the core is considered to form preferential
water passages for the movement of fluid, which are similar to fractures in consolidated rock. A rock
fractures K value is proportional to the square of its aperture (Snow 1969; Domenico and Schwartz 1998;
Singhal and Gupta 1999). A further assumption is that the water passage width in each unit of the low
packing part is linearly related to the length fraction of each packing level. In addition, the proportionality
coefficients are assumed to differ between packing levels III and IV. Consequently, the K value for the low
packing part takes a simple form:
K LP C i L i ,
2
(4.5)
where KLP is the hydraulic conductivity of the low packing part and Ci is a dimensionless proportionality
coefficient, which takes a different value for packing level III (index i = 3 and level IV (index i = 4).
(3) An equivalent horizontal hydraulic conductivity K per unit depth, composed of a sequence of
different packing levels, is estimated as a weighted arithmetic mean, specifically, the sum of the products of
each packing levels individual K value and length fraction. Accordingly, K is calculated by the following
equation:
4
(4.6)
The constants CHP, Ci, and m in Eq. (4.6) are simultaneously determined under the hypothesis that the
estimated K values correspond to the slug test results at the same depth. j is the index of boundary
between high packing and low packing parts. The agreement between estimated and actual K values is
then assessed by using a least squares method to minimize the root mean square error (RMSE). The RMSE
value is calculated as the sum of the residuals between the common logarithms Y of K and the
logarithms Y of K measured by the slug tests at the same depth:
n
(4.7)
where ReY is the residual log-conductivity and n is the sample number (= 32). Logarithmic transformations
93
are used to avoid a few extremely permeable values from considerably affecting the RMSE results. A
nonlinear optimization tool, the MS Excel Solver, was used to perform the optimization. Since d is not
empirically known, it is determined by repeatedly optimizing using various cumulative weight diameters;
d5, d10, d15, d20, d30, d40, d50, d60, d70, d80 and d90 (mm).
4.2.3
The core properties profiles in the sampling points, BW3, BW5 and BW7 were transformed into those
of K by using Equation (4.6). However, the optimal equation (4.6) has inevitable uncertainty because
the parameter optimization is conducted by log-transfomation in Equation (4.7). If the residuals between
Y and Y (ReY) are only no more than one, the confidence interval of K was extended by several to
10-fold when the log transformation was inverted. The ReY resulted from various factors: the sampling
limitations of using small diameters and volumes, the over- or under-estimation of the packing level, the
applicability of the slug test method to the test conditions, and the simplifications made to generate the
relation in Eq. (4.6).
To address the uncertainty in Eq. (4.6), moving average method was applied to the profiles of
Y in the sampling points. The moving average of Y (Y MA ) is calculated through the following equation:
nMA
nMA
Y MA Y i log10K i ,
(4.8)
where Y i is the common logarithm of estimated conductivity K i for ranging from 1 to nMA. nMA is the
total number of K values used for the moving average; in the present case, nMA corresponds to the
average interval written in meter units, because K i was calculated per unit depth. The average interval
must be determined carefully since the uncertainty in Y MA decreases when the average interval increases
under a stationary condition. Conversely, Y MA is influenced by a spatial trend (which is empirically
unknown), when the average interval is relative larger than the trend. Here, an average interval of nMA = 5
m was applied. Y MA was thus the average of five Y values; the Y value at target depth and two values at
depths above and below the target depth. In traditional statistics, the 95% confidence interval of Y MA was
obtained as [Y MA 1.96
nMA , Y MA + 1.96
94
4.3
.3
4.3.1
The some results of grain size analysis are shown in Figure 4.7. All of the samples were poorly sorted
as there was a wide range of grain size present. Less than 5% of the samples consisted of silt and clay, and
more than 1040% consisted of gravels. Figure 4.7 showed no obvious trend among different locations,
for example a fining trend in the downward direction from the upper point (BW6) to the lower (BW1).
100
BW01-1(GL-3.04.0m)
BW02-1(GL-3.04.0m)
BW03-1(GL-3.04.0m)
BW04-1(GL-4.05.0m)
BW05-1(GL-3.04.0m)
BW06-1(GL-3.04.0m)
90
80
70
60
50
40
30
20
10
0
0.001
1E-3
0.01
0.005
Clay
0.1
1
Grain size(mm)
0.075 0.425
2
Sand
Silt
Fine
Coarse
10
100
19
75
Gravel
Fine
Coarse
Figure 4.7 Grain size distribution curves of undisturbed cores sampled at BW16
Figure 4.8 shows relations between two grain sizes (d10 and d50) and slug test results obtained in
the sampling depths. There were also no obvious relations in both plots, indicating that K in the
coarse gravel deposits could not be formulated only by using the traditional equations, Eq. (4.3) and
(4.4). In other words, K in the coarse sediments is not a function of only total porosity; if total
porosity is even small but open voids are connective between gravels, the coarse gravel deposits has
high permeability. It was needed to indicate another index which reflected the open pore structures
in the gravel deposits.
95
-2
10
-3
10
-4
10
-5
10
K=1.8810 d10
-4
1.70
R =0.22
-6
10
-1
10
10
10% diameter on the grain size diagram (mm)
10
-2
10
-3
10
-4
10
-5
10
-6
K=1.2210 d50
1.29
R =0.17
-6
10
10
50% diameter on the grain size diagram(mm)
10
Figure 4.8 Relations between grain size (d10 and d50) and hydraulic conductivity
obtained by slug tests in the sampling depths.
96
4.3.2
The first study (Sakata et al. 2011) determined the index j (j= 1-4) and effective grain size de (de=
d5, d10, d15, d20, d30, d40, d50, d60, d70, d80 and d90) through a comparison of RMSE values obtained by Eq.
(4.7). The determination was conducted employing only the data at BW1 to BW6, and the K values
was not compensated to the values of constant temperature of 25 C.
The minimum RMSE value was obtained in the case of j=1 and de=d20 (Model 1 in Table 4.2),
however, the case of j=2 and de=d20 (Model 2) was most optimal as followings: 1) the degree of
freedom in Model 2 was smaller than that in Model 1; 2) the constant C2 in Model 1 was enough small
to be ignored, and 3) the other constants, C3 and C4, were almost equal in the cases.
Table 4.2 Results of parameter optimization using undisturbed cores and slug
tests at BW1-6
Model 1: j =1,C m=C 1C 2,C 3,C 4
d 10
RMSE
m (-)
C m (1/m2s)
C 2 (1/m2s)
C 3 (1/m2s)
C 4 (1/m2s)
0.97
2.9
28,800
0.00048
0.010
1.1
d 20
0.68
3.2
8,642
0.00038
0.010
0.65
d 30
d 40
d 50
0.81
2.2
2.0
0.00037
0.011
0.71
0.91
1.5
0.014
0.00026
0.012
0.78
0.96
2.5
0.19
0.00030
0.012
0.75
d 50
d 10
d 20
d 30
d 40
0.99
2.3
516
0.0098
1.0
0.83
2.2
30
0.0093
0.69
0.89
1.6
0.12
0.010
0.74
0.90
1.5
0.013
0.012
0.78
d 30
d 40
d 50
1.29
1.0
0.012
0.91
1.30
1.0
0.0049
0.91
1.37
1.2
0.0046
0.94
d 30
d 40
d 50
1.89
1.0
0.031
2.01
0.7
0.0024
0.95
2.4
0.16
0.013
0.75
d 10
RMSE
m (-)
C m (1/m2s)
C 4 (1/m2s)
1.35
1.3
0.86
1.1
d 20
1.24
1.4
0.40
0.86
d 10
RMSE
m (-)
C m (1/m2s)
2.05
0.6
0.011
d 20
1.86
1.2
0.25
1.86
2.1
0.30
97
Following the sampling at BW7, the next study (Sakata and Ikeda 2012b) calculated the optimal
parameters again. The analysis data and results are listed in Table 4.3. In the calculation, the K values
were converted into those at a constant temperature of 25 C for heat transport simulation in Chapter 5.
The optimization resulted that the minimum RMSE value was obtained as:
K 6.89(d20/1,000)1.9 L1 L2 0.0167L3 1.87L4 .
3
(4.9)
A scatter plot of estimated K versus measured K values is given in Figure 4.4. The coefficient of
determination (R2) between Y and Y is relatively high at 0.80.
98
Table 4.3 Analysis data and optimization results to determine the relation between
slug tests and gravel cores in the Toyohira River alluvial fan.
Sample
number
Point
identifier
Sampling and
test depth (m)
Log-conductivity
L1+L2 b
L3 b
L4 b
Yc
Yd
ReY e
0.91
0.00
0.09
-2.62
-2.86
0.24
BW1
910
BW2
56
12
0.29
0.53
0.18
-2.05
-1.86
-0.19
89
1.6
0.80
0.18
0.02
-3.16
-3.86
0.70
34
0.83
0.50
0.40
0.10
-2.55
-2.53
-0.02
910
3.7
0.43
0.42
0.15
-2.76
-2.12
-0.64
1920
2.4
1.00
0.00
0.00
-4.09
-4.12
0.03
2930
8.6
0.66
0.25
0.09
-2.50
-2.66
0.16
3132
0.70
0.91
0.09
0.00
-5.15
-4.73
-0.42
3940
4.2
0.81
0.19
0.00
-3.58
-3.54
-0.04
10
4950
1.0
1.00
0.00
0.00
-5.35
-4.86
-0.49
11
5960
3.0
1.00
0.00
0.00
-4.14
-3.94
-0.20
BW4
56
1.0
0.82
0.09
0.09
-3.22
-2.86
-0.36
89
0.63
0.60
0.40
0.00
-3.51
-2.97
-0.54
BW5
34
2.2
0.71
0.29
0.00
-3.00
-3.35
0.35
3
4
12
BW3
13
14
0.85
15
910
2.0
1.00
0.00
0.00
-4.80
-4.29
-0.51
16
1920
3.3
0.85
0.15
0.00
-3.65
-3.77
0.12
17
2930
1.5
0.93
0.00
0.07
-3.67
-3.18
-0.49
18
3940
0.71
1.00
0.00
0.00
-4.64
-5.15
0.51
19
5960
2.1
1.00
0.00
0.00
-4.48
-4.24
-0.24
20
6970
0.86
0.90
0.10
0.00
-5.03
-4.58
-0.45
8990
1.43
0.92
0.08
0.00
-4.66
-4.47
-0.19
67
2.1
1.00
0.00
0.00
-4.56
-4.26
-0.30
23
910
5.2
0.93
0.04
0.03
-3.03
-3.46
0.43
24
1819
4.4
0.84
0.11
0.05
-3.63
-3.35
-0.28
1011
2.8
1.00
0.00
0.00
-3.79
-4.01
0.22
26
1415
1.3
1.00
0.00
0.00
-3.76
-4.65
0.89
27
1819
1.0
0.77
0.13
0.10
-2.28
-2.72
0.44
28
2324
0.81
0.94
0.00
0.06
-2.99
-3.38
0.39
29
2930
0.55
0.96
0.00
0.04
-3.55
-3.91
0.36
30
3334
2.2
0.88
0.03
0.09
-2.82
-2.85
0.03
31
3738
1.1
0.82
0.18
0.00
-3.43
-3.96
0.53
32
4344
1.1
1.00
0.00
0.00
-4.85
-4.81
-0.04
21
22
25
BW6
BW7
Mean
0.00
Variance
0.16
99
The relation between estimated K and d20 for cases of L3 and L4 is shown in Figure 4.10. When L3 <
0.05 m/m and L4 is rarely observed on the centimeter scale (i.e., the gravel core in effect contains only a
high packing part, K varies from 5 10-6 to 2 10-3 m/s. Here, K is governed by only the grain size, and
is independent of core packing level. When L3 > 0.5 m/m or L4 > 0.1 m/m (i.e., the occasional case that
occurs near the surface (Figure 4.6)), K is of the order of 1 10-3 m/s for all grain sizes. Such a high
conductivity regardless of the length scale for the low packing part indicates that preferential water
passages may exist at the sampling depth. When 0.05 L3 0.5 m/m and L4 is of the order of 1 10-2 m/m
(i.e., the intermediate case), either the grain size or length fraction of the packing levels has a strong affect
on K . The limit on the effective grain size that governs its effect on K is roughly d20 =-2 (- = 4) mm,
where - denotes the base-2 logarithm of grain size (mm).
Estimating K from Eq. (4.9) results in it attaining a high value even if the low packing part is fairly
concentrated. This hydraulic feature corresponds to that of open framework gravel (OFG), which is usually
observed in outcrops or trenches, and is considered to be the most remarkable hydro-face due to its high
permeability. OFG has a distribution that is only centimeters or decimeters thick, but its K value is of the
order of 1 10-3 to 1 10-2 m/s, considerably greater than the value for the surrounding layers (Jussel et al.
1994; Heinz 2003; Lunt et al. 2004; Zappa et al. 2006; Ferreira et al. 2010; dellArciprete et al. 2012).
Figure 4.10 Curves relating K calculated by Eq. (4.9) and effective grain size
diameter d20 = log2 d20 (mm) for several length fractions of low
packing part L3 or L4 (m/m)
100
ReY ranged from -0.64 to 0.89 with a variance of 0.16, as shown in Table 4.3. These statistical values
of the residuals were not necessarily ignored because the confidence interval of K was extended by 3- to
10-fold or greater when the log transformation was inverted. Considering an assumptions that Y was
normally distributed and K was log-normally distributed (Domenico and Schwartz 1998; ASCE 2008),
and that ReY was also normally distributed with a constant variance of 2 = 0.16, specifically, ReY was
assumed to occur randomly with no spatial correlation and with a constant variance. A 95% confidence
interval for Y was obtained under the above assumptions as [Y 1.96 , Y + 1.96 ] (i.e., [Y 0.78,
Y + 0.78] in other words, [ K 6 .0 , 6.0K ]. The interval is approximately equal to a single order of
magnitude and so K was not quantitatively valid in this case. On the other hand, the 95% confidence
interval of Y MA was obtained as [Y MA 0.35, Y MA + 0.35], and was transformed to [ K 2.2 , 2.2K ].
This range was considered to be sufficient for quantitative discussion of the vertical trend.
101
4.3.3
The results for BW5, BW3 and BW7 are shown in Figure 4.11. From left to right, the results for each
sampling point show changes with depth of the geologic column, effective grain size diameters, length
fractions of the low packing part and log-conductivities: Y measured by slug tests, Y estimated by Eq.
(4.9) and Y MA over 5-m intervals. Gaps in the profiles indicate data rejected or not obtained at depths
where cores are composed mainly of fine sediments.
The geologic columns at BW5 and BW3 do not show migration sequences of the gravel deposits, which
are generally a characteristic of alluvial fans rather than meandering river sequences. A majority of the
intercalating sandy layers are 1 m in thickness, and are rarely seen in consecutive horizontal positions
among the wells. Only monotonic gravel successions therefore show no obvious hydrogeologic boundary
in the gravel aquifer (e.g., the boundary between Holocene no. II and Pleistocene no. III aquifers). In
contrast, the gravel deposits at BW7 lie below an alternation of fine sediments, which corresponds to the
upper edge of the Holocene no. I aquifer. Additionally, the intercalating layers in the gravel deposits are
frequently distributed with even finer sediments. In particular, volcanic ashes are typically interbedded at
depths of ~1920 and 4143 m, corresponding to the reworked deposits of Shikotsu pumice flows
(~40,000 yr) before present (B.P.) and of Toya ash falls (~110,000 yr B.P.), respectively.
d20 varies in each well, ranging widely from -1 to 4 (-) (i.e., ~0.5 to 16 mm). Moreover, the deviations
in adjacent data in the profiles are often greater than 1 to 2 (-). This large fluctuation is considered to mask
the spatial trend, for example, a down-fan fining trend from the mid-fan to fan-toe. Such fluctuations
probably arise from sampling errors due to various sources (as already described), as well as from
sedimentary heterogeneity. Here, no distinct difference is found between the grain size variations of BW5,
BW3 and BW7, but a decreasing trend with depth is observed only at BW5, although this is not obvious
due to the fluctuation.
The length fractions of the low packing part, L3 and L4, show a clear decreasing trend with depth at BW5
and BW3. L4, in particular, decreases rapidly with depth, and has almost vanished by a depth of ~30 m. A
somewhat decreasing trend is also seen in L3 such that its values appear to be relatively random. In contrast,
a specific trend is not seen for L3 or L4 at BW7. The main reason for this difference in trends is that in the
mid-fan, the low packing part accumulates in the gravel deposits near the surface, whereas in the fan-toe,
102
the late Holocene fine sediments cover the surface with a thickness of >10 m. Another reason is considered
to be the development process of OFG, the hydraulic features of which are equal to those of the low
packing part. Thick OFG deposits occur due to high bedforms and large amounts of aggradation (Lunt and
Bridge 2007), and form in the mid-fan, but not in the fan-toe.
Y values are measured by slug tests at several points in each well. As a result, Y ranges from -2 to -6, that
is, four orders of magnitude of K (m/s), by inverting the log transformation. This large range of
measurements is considered as being reflective of the geological heterogeneity in the gravel deposits.
However, a decreasing trend in Y is found at BW5 and BW3. Conversely, Y values in each profile
fluctuate even more widely, with an amplitude of >12, which corresponds to one to two orders of
magnitude of K . The greater fluctuation in Y is probably a result of not only the heterogeneity but also the
estimation errors. Therefore, Y values are not used to determine the exhibited trend. Instead, Y MA values
in each profile vary more smoothly, and thus reveal the changes with depth. Decreasing trends in Y MA are
evident at BW5 and BW3, and are observed only above depths of ~30 m, which corresponds to the
maximum depth at which L4 values are observed. Furthermore, the moving average below this depth is -4,
showing that an approximately stationary field may exist. In contrast, a vertical change with depth is not
obvious at BW7, where trends are not observed for L3 or L4. Thus, the depth dependence of K in the fan is
understood from the vertical distribution of the packing levels in the undisturbed cores.
103
Figure 4.11 Core property and hydraulic conductivity results at a. BW5, b. BW3 and c. BW7. Each results contains, from left to right, the geologic column;
effective grain size diameters, = log2 d20 (mm); length fractions of low packing part, L3 (gray bars) and L4 (dark gray bars) (m/m); and
common logarithms of the hydraulic conductivity (m/s). In the log-conductivity profiles, solid circles denote Y values measured by slug tests,
open squares denote Y values estimated by Eq. (4.9) and solid lines denote moving average Y MA values over 5-m intervals. All data are
obtained at unit depth. Gaps represent depths at which data were rejected or not obtained
104
4.3.4
A linear regression analysis is applied between Y MA and depth. Here, the depth data correspond to the
midpoints of each average interval and the Y MA data are taken above depths of 40 mthe maximum
depth of the decreasing trend, 30 m, plus an additional 10 m. A scatter plot showing the regression results is
given in Figure 4.12. The regression lines for BW3 and BW5 have approximately equal slopes and
intercept values; the slopes are 0.041 and 0.052 (average: 0.047), respectively, and the intercepts are -2.5
and -2.2 (average: -2.3). In addition, R2 values of >0.7 are obtained. A depth dependence of K is often
represented as an exponential function:
K K 0 exp( AZ ), (4.10)
where K 0 (m/s) is the K value at the ground surface, which is equal to 1 10-2.3 m/s for the current case;
A (m-1) is a decay exponent; and Z (m) is the depth below the surface. The decay exponent is determined
by dividing the average slope of 0.047 by log 10 = 2.3. Hence, A = 0.11 (m-1). The exponent equation for
the mid-fan is therefore
K K 0 exp( AZ ) 10 2.3 exp(0.11Z ).
(4.11)
Figure 4.12 Scatter plot of Y MA values at BW5, BW3 and BW7 with associated
regression lines. Solid squares denote Y MA values at BW5, open circles
denote values at BW3 and solid triangles denote values at BW7.
Averages are plotted at mid-depths of 5-m intervals. A regression
analysis between Y MA and depth is conducted in each well using data
above a depth of 40 m
105
The size of the decay exponent here should be noted. For example, an increase in depth of 1 m
corresponds to a ~10% decrease in K. Explicitly, K at a depth of 10 or 30 m is equivalent to approximately
1/3 or 1/25 of K 0 at the ground surface. The exponents in other unconsolidated gravel deposits are
regrettably not obtained in this study; thus, previous values found for various consolidated rocks are used
for comparison. These values differ greatly at each site, and are of the order of 1 10-2 to 1 10 -4 m-1, as
described before. The exponents for consolidated rocks are therefore not smaller than 1/10 to 1/1000 of the
gravel deposit value. Conversely, the regression line for BW7 has an R2 value of only 0.13. This small
value implies that a vertical trend is not evident in the fan-toe (although only one profile is created).
Specifically, the depth dependence of K is not common throughout the fan, but is restricted to the upper fan,
where gravel deposits of the low packing part accumulate below the ground surface.
The typical characteristics in the vertical trendthe large exponent A in the mid-fan and the fan-apex,
and the lack of a trend in the fan-toeare considered as follows. One simple and most probable argument
is that the depth dependence at BW5 and BW3 is a unique stratigraphic trend (i.e., coarsening-upward
trend) due to progradation in the Holocene fan. Stationarity below a depth of ~30 m may indicate different
(lower energy and slower) depositional environments such as braided river system. The lack of a trend at
BW7 indicates that the location is adjacent to, but not in, the alluvial fan depositional system. However, the
interpretation based on only stratigraphic sequences remains slightly problematic. The low packing parts in
the gravel cores are rarely observed at deep depths, although the grain sizes at deep depths are often large,
no smaller than those in at shallow depths. Another argument is specific post-depositional processes in
shallow alluvial deposits, other than the often discussed diagenetic and metamorphic processes. In shallow
burial, crushing probably has a small effect on the porosity reduction, and cementation such as
groundwater calcretization is not seen in the cores. An understanding of the post-depositional processes in
the shallow gravel deposits is most likely needed to obtain other information such as the stability of fine
sediments in the openings between framework components. If the gravel deposits are loosely packed and
the fine sediments are unstable, the open pore structures that correspond to the low packing parts in the
cores are theoretically more densely filled due to the movement of the fine interstitial sediments through
increased fluid pressure with buried depth. To our knowledge, the mechanism of the post-depositional
transition from loose to high packing in alluvial coarse mixtures has not been examined. Although the
factor in the trend, whether depositional or post-depositional processes, has not been determined, this field
106
investigation reveals a unique vertical trend in the investigated alluvial fan, and the depth dependence is
well-formulated as the exponential model used for the analysis and numerical modeling of groundwater
systems.
4.3.5
Aquifer structure
A longitudinal cross section is generated by using the geologic columns and Y MA profiles, as shown in
Figure 4.13. This cross section extends from the fan-apex 2-km upstream of BW5 to the fan-toe at BW7.
The two dotted lines are the boundaries determined from only the stratigraphic information. The lower
dotted line is related to the boundary between Pleistocene nos. III and IV aquifers, which is distributed
throughout the cross section at an elevation of around -50 meters above sea level (m asl) according to Hu et
al. (2010). The upper dotted line represents the boundary between Holocene nos. I and II aquifers, which is
determined to be ~10 m asl at BW7. These dotted lines are relatively distinguishable because of the sudden
stratigraphic change between the gravel fan deposits and the fine fluvial deposits. However, other
boundaries of the gravel deposits such as between Holocene no. II and Pleistocene no. III aquifers are hard
to specify among the monotonic stratigraphic columns.
Two Y MA boundaries are also traceable among the profiles: a lower broken line with Y MA = -4 and an
upper broken line with Y MA = -3. The lower boundary is below a depth of ~30 m from the ground surface,
and is at the same elevation as the hydrogeologic boundary between Holocene no. II and Pleistocene no. III
aquifers (Hu et al. 2010b). This correspondence indicates that the Pleistocene no. III aquifer has only weak
K depth dependence, and the permeability in the aquifer is assumed to be stationary at around the
expectation of K = 10-4 m/s. In contrast, the upper boundary of Y MA divides the Holocene no. II aquifer
into two further aquifers: nos. IIa (upper) and IIb (lower). The Holocene no. IIa aquifer is ~10 m thick near
the surface, and Y MA > -3 here (i.e., K = 1 10-3 m/s). Such high values of K are similar to those found
in previous pumping test results: pumping tests were conducted in shallow wells near BW5 and BW3, and
obtained an average hydraulic conductivity of K 2 10-3 m/s (Hokkaido Regional Development Bureau
2008). The Holocene no. IIa aquifer is directly connected to the river bed, and the surface water is actively
infiltrated through the aquifer due to its high permeability and vertical hydraulic gradients. Thus, the
Holocene no. IIa aquifer along the river is of importance as an infiltrative aquifer. However, if the aquifer is
at a distance of several hundred meters from the river, it is almost unsaturated since the groundwater
107
108
Figure 4.13 Longitudinal cross section showing hydrogeologic boundaries in the fan. The columns and profiles for BW5, BW3 and BW7 are the same as those
Fig. 7. Inverted open triangles beside each column denote groundwater heads measured during drilling at the first (upper) and last (lower) depths,
and indicate vertical hydraulic gradients in the mid-fan at BW5 and BW3. Dashed lines denote Y MA boundaries inferred from the profiles: Y MA = -3
(upper) and Y MA = -4 (lower). Dotted lines are boundaries inferred from geologic columns and Hu et al. (2010). Classification of aquifers, nos. I, II,
III and IV, are taken from Hu et al. (2010b). Aquifer no. II is further subdivided into nos. IIa and IIb.
109
4.4
Conclusions
The depth dependence of the hydraulic conductivity K in the Toyohira River alluvial fan was
determined, and the trend was represented by an empirical exponential equation. The proposed method
consisted of the following steps.
(1) Relatively undisturbed and sequential gravel cores were sampled by using improved tube samplers.
(2) The packing in gravel cores was categorized qualitatively into four levels, and then the length
fraction of each packing level was measured per unit depth. After these measurements, the effective grain
size (in the current case this was d20) was obtained by sieving.
(3) The relation between the slug tests and core properties was established through optimization.
(4) Core properties profiles were transformed into those of the estimated hydraulic conductivity K by
using the established relation.
(5) A moving average method was applied to eliminate errors. In the present case, an average interval
of 5 m was used.
(6) A linear regression analysis revealed the depth dependence of K in the mid-fan, and the decay
exponent was estimated.
The depth dependence of K was shown at the sampling points BW5 and BW3, and the decreasing
trend had a decay exponent of A = 0.11 (m-1), which is 10- to 1000-fold that for consolidated rock in the
literature. Conversely, at BW7 in the fan-toe, a vertical trend was not observed. A longitudinal cross
section was further generated by using the moving average profiles. Moving average boundaries of Y MA
= -4 and Y MA = -3 classified the gravel aquifer without migration succession. As a result, the
hydrogeologic structures were understood as a high infiltrative aquifer (Holocene no. IIa), a
depth-dependent and heterogeneous aquifer (Holocene no. IIb) and a stationary permeable aquifer
(Pleistocene no. III).
Several problems remained unsolved in this study. The relation between the slug tests and core
properties has not been sufficiently verified either theoretically or experimentally. Moreover, the obtained
relation was applied only in the fan, and individual relations must be established at each site. The
sampling error of the sampler has not been assessed experimentally (e.g., by comparison with bulk
sampling). The scale at which permeability was observed in this study (i.e., unit depth is not necessarily
110
appropriate at other sites), and the employed scale must be discussed in terms of each hydrogeologic unit
(Anderson 1989). Furthermore, the proposed method provided only vertical information at the sampling
points. Thus, for three-dimensional groundwater modeling, horizontal information must be acquired by
other methods such as outcrop analyses, geophysics surveys or deterministic depositional models. In spite
of these issues, the proposed field method is expected to be useful for gaining greater quantitative insight
into the depth dependence of the permeability in unconsolidated gravel deposits.
111
Introduction
Any analytical methods to solve the governing equation of groundwater are limited to
particular problems in which all of the region of flow, boundary conditions, and geologic
configuration are simple and regular. Numerical modeling and simulation are versatile for such
complex and actual conditions in the geologic settings. Geostatistical approaches also contribute
to realize spatial variability of hydraulic properties such as hydraulic conductivity K. However,
the approaches need typical information about global trends in trending heterogeneity of the
alluvial coarse deposits. The previous chapter indicates the depth-dependence of K in the
fan-gravel deposits as the exponential function, of which exponent decay is 10- to 1000- orders
of magnitude larger than that in consolidated rocks. This chapter currently realizes trending
heterogeneity in the gravel deposits by incorporating the global trend and stochastic simulation,
and allows assessment of its importance in groundwater modeling.
Hydraulic conductivity K in the gravel deposits is represented as a sum of a global trend
component and a residual component as the groundwater table elevation is mapped in Chapter 2.
The global trend is divided into those in the vertical and horizontal directions. The vertical trend
has been obtained as the exponential function in the previous chapter, and then the horizontal
trend is estimated as a linear regression model of coordinates in this chapter. The residual
component of K offers its spatial variability as an anisotropic variogram model. 100 realizations
of residuals are produced using sequential Gaussian simulation, and the multiple realizations of K
are obtained by summing the residual realizations with the trend values at given points in the
analysis domain. 13 study cases each of which includes 100 realizations are prepared as a
comparison study, i.e., stationary or trending heterogeneity, isotropy or anisotropy of each cell,
River conditions (width and vertical hydraulic conductivity), spatial variability of residual,
temperature-dependence of fluids or independence, upscaling affect on simulation results.
Solute transport simulation is more susceptible to geologic heterogeneity than only flow
simulation is. Hydraulic head contours usually look smooth as a result of a pressure balance,
while solute concentration contours have sharp fronts reflected to actual flow of groundwater, not
112
Darcys flow. However, the coupled model involves more difficulties than the uncoupled flow
model: 1) scarce measurements of solute concentrations, 2) higher uncertainty in parameters of
fluid and aquifer (e.g., diffusivity), and 3) numerical errors associated with spatial and temporal
discretization. For example, a high-resolution model (e.g., grid spacing of 1 m) is needed to
obtain detailed, accurate results to solve for the solute concentration, but excessive resolution is
inappropriate for larger-scale models (e.g., a three-dimensional basin model) due to
computational limitations.
The first and second problems are addressed by using heat as a groundwater tracer by
exploiting the similarity between temperature transport and solute transport: compared with other
solute components, heat can be measured with greater ease and accuracy, and the representative
values of thermal parameters are obtained from the literature.
Also, most of thermal properties in various materials and fluids are described in literature.
Especially, heat has been utilized for analyzing the movement of groundwater and its interaction
with surface water in many previous studies (Lapham 1989; Stonestrom and Constanz 2003;
Anderson 2005; Constanz 2008; Constanz et al. 2008; Healy 2010; Anderson 2010). Temperature
of surface water flowing in the open channels is variable diurnally and seasonally, while that of
groundwater beneath the streambed is more invariable. Many field studies have been performed
based on the difference of temporal variation between surface water and groundwater. In
particular, temporal fluctuations (envelope) of groundwater temperature are often observed in the
wells below the water bodies, and 1D analytical model is utilized to estimate the properties of
geologic material and the velocities of vertical groundwater flow (Taniguchi 1993; Taniguchi
1999; Vandenbohede and Lebbe 2010). However, numerical simulation of heat transport in a
groundwater system has been rarely performed due to the computational expensiveness. K in a
geologic material consists of fluid density and viscosity in addition to intrinsic permeability, such
that K is inherently dependent on groundwater temperature. In most conventional models, the
effect of temperature is ignored. Recent studies, however, suggest the importance of temperature
dependence with respect to solution robustness in groundwater modeling (Doppler et al. 2007;
Ma and Zheng 2010; Engeler et al. 2011). If the temperature of river water varies between about
zero in winter to over 20 C in summer, the lowest K value of sediments around the river is about
113
a half of the highest. More rigorous solutions in groundwater flow and solute transport are
obtained by using advanced models, in which density and viscosity of fluids are variable as
functions of temperature (Ma and Zheng 2010; Engeler 2011). A computer program of variable
density and viscosity of fluids flowing in porous media, SEAWAT (Guo and Bennett 1998; Guo
and Langevin 2002; Langevin et al, 2003; Thorne et al. 2006), is used in this study. The program
used, SEAWAT, couples the flow program MODFLOW (Harbaugh et al. 2000) and the solute
transport program MT3DMS (Zheng and Wang 1999). To simulate heat transport within the text
of the SEAWAT framework, one of the MT3DMS species is used to represent temperature, and
the effects of changing density and viscosity can be considered (Ma and Zheng 2010). SEAWAT
also calculates groundwater flow and temperature transport with greater accuracy by using
iterative processes. The third problem is also addressed by using a high-resolution grid model.
Previous study (Weismann and Fogg 1999; Weismann et al. 1999) suggested that a
high-resolution numerical model of groundwater flow and solute transport is needed to assess the
effect of complex heterogeneity in groundwater mixing and tracer concentration. In this
dissertation, a high-resolution grid model of 1 m grid-spacing is designed to satisfy the criterion.
A block-averaging model consisting of 5 m square cells is also prepared for a comparison with
the high-resolution model. Upscaling of a high-resolution model to a basin-scale model is a
long-standing problem in groundwater modeling (Deutsch 2002; Zhang et al. 2006; Zhang et al.
2010). This is because K depends on scale, that is, microscale (e.g., drilling cores; several meters
or less), mesoscale (e.g., facies; 10 to 100 m), and macroscale (basin scale, 100 m or more). The
discrepancy between the different discretized models shows a direction of upscaling in
groundwater modeling of alluvial coarse deposits.
5.2
5.2.1
Hydraulic head and groundwater temperature were measured in the observation wells located
along the off-stream transect near the Horohira-Bridge (KP16.6) in the distinct losing section
(KP15.517, Chapter 3). Four observation wells of different locations and depths were installed
by the Hokkaido Regional Development Bureau between March and June 2008; deep and
114
shallow wells near the river channel (BW31 and BW32); a shallow well leftward from the
channel (BW34); and a shallow well rightward from the channel (BW35). The observation
wells, BW31 and BW32, correspond to the wells, OW15D and OW 15S in Table 1.1,
respectively. The details of wells are summarized in Table 5.1, and the locations are shown in
Figure 5.1.
The automatically-recorded sensors of two different types, U20-001-01 HOBO pressure
transducer (an accuracy of +/0.05 cm, HOBO Inc., USA) and U22-001 HOBO thermal sensor
(an accuracy of +/0.02C), were inserted into the observation wells. The hourly measurements
were carried out from 10 June 2011. The hourly stages of river stage in the Kariki and Moiwa
gaging stations were publicly available from the website Water Information System of the
Ministry of Land, Infrastructure and Transport, Japan. The river water temperatures in the
Horohira-Bridge were also given in the same website, but the measurement regrettably stopped at
21 March 2011.
Temperature logging in the deep well BW31 was also performed to obtain seasonal profiles
of groundwater temperature in the focused recharge zone. The temperature profiles obtained were
those at 9 June 2010, 9 August 2010, and 3 December 2010. The logging was performed per 0.1
m above a depth of 30 m, and per 0.5 m below the depth using a thermistor made of platinum
resistance (an accuracy of +/0.01C).
Table 5.1 Observation wells in the off-stream transect (Figure 5.1). X denotes
horizontal distance from the center of river channel: SC denotes well
screen depth: H denotes groundwater head at 10 June 2011: Z
denotes depth of temperature sensor(s). BW33 is omitted because
of the same location and depth with BW32
Well
X
SC
H
Z
BW03-1
-62
62 - 64
20.93
15, 30, 60
BW03-2
-66
2.5 - 7.5
28.67
5
BW03-4
-140
9.5 - 11.5
28.69
11
BW03-5
50
3-5
28.27
5
115
116
Figure 5.3 a) pressure transducer and b) thermal sensor used for this study
5.2.2
Non stationary problem that faces geostatistical approaches in the alluvial coarse deposits was
addressed using the methodology of regression kriging (Hengel 2007), as well as the mapping of
groundwater table elevation in Chapter 2. The regression kriging process requires the assumption
that regionalized variable (hydraulic conductivity in this case) is a sum of a global component
and a residual component. The vertical trend of K was formulated as the exponential function in
117
Chapter 4. The logarithmic transformation is often used in groundwater study because K values
are positive and relatively many small values exit compared to larger values (e.g. Freeze, 1979;
Anderson 1995; Domenico and Schwartz 1998; ASCE 2008). Thus, the vertical trend of common
logarithm of K is mathematically represented as:
logK 0(x)z=0= logK + Az/ln(10)
logK 0(x)z=0= logK + Az0
at zz0
at z> z0
(5.1)
(5.2)
where K 0(x) z=0 is hydraulic conductivity at the ground surface (depth z=0) of the x position: A is
the exponent decay of K with depth (A= 0.11m-1 in this case): z0 is the maximum depth of the
depth-dependence (z0= 30 m in this case).
The horizontal trend of common logarithm of K was then estimated by formulating a linear
regression model of coordinates X and Y (JGD 2000) as:
logK0(x)z=0= aX+ bY+ c
(5.3)
K0 at the ground surface was obtained by translating the measured logK using Equation (5.1)
and (5.2). Averaged logK0 value was applied when in-situ tests were obtained at different depths
in the same location. The constants a, b and c in Equation (5.3) were estimated by the
straightforward linear regression method. Consequently the global trend component M(x) in
common logarithm of K is given as:
M(x)= Mh+ Mz= (aX+ bY+ c)+ Az/ln(10) at 0< z z0
(5.4)
(5.5)
where M(x) is the trend component at any points x in the domain: Mh is the horizontal trend on
the ground surface,: Mz is the vertical trend: A is the exponent decay (0.11 in this case): z0 is the
maximum depth of depth-dependence of K (z0 = 30m in this study): zbottom is the bottom depth of
the model (e.g, zbottom= 64 m in BW31).
The residual component at the measurement points xi, where i denotes the index of
measurement points, were calculated as:
Re(xi)= logK(xi) M(xi)
(5.6)
where Re(xi) is the residuals at the in-situ test points xi. In-situ measurements of K for this
analysis were summarized in Table 5.2. Table 5.2 also shows residual Re(xi) calculated by Eq.
(5.4), (5.5), and (5.6).
118
5.2.3
Sequential Gaussian simulation (SGS) is the most powerful and popular among various
simulation processes for realizing geologic heterogeneity (Koltermann and Gorelick 1996; Lee et
al. 2007). SGS requires an assumption that the probable density function (pdf) of target (residual
in this case) attribute is Gaussian that is defined by only mean and variance. Under the
assumption, the statistical hypothesis of second-ordinary stationarity is simultaneously
established. In sequential Gaussian algorithms, grid nodes at which values have not yet been
assigned are selected at random, and one value is drawn a random on the local conditional
probability distribution. The mean and variance in local conditional probability distribution is
estimated using kriging estimation (e.g., ordinary kriging) of nearby data at previously simulated
grid and conditional grid. The process is repeatedly performed until simulation values at all
desired nodes are determined.
This analysis produced 100 realizations of residuals, Re(x), at a regular-grid of 1 m in the
cross-sectional area using a free Fortran program SGSIM in GSLIB (Deutsch and Journel 1998).
The theroretical variogram of Re(x) required in SGS was an anisotropic spherical model as:
3 h 1 h 3
2 a 2 a
h B CB
(5.7)
where B was the nugget effect, C is the partial sill, a is the range, and h is the separation distance.
The lateral- and vertical- variograms of residuals Re (x) were produced from 67 in-situ
measurements of K in the fan as shown in Table 5.2. In this study, the experimental and
theoretical variograms of Re(xi) were produced by using the software Surfer (Goldensoftware,
Inc.). These parameters a, B, and C were also determined through the non-linear regression
method in Surfer. The products of residuals were then translated to common logarithm of K as:
logK(x)= M(x)+ Re(x)
(5.8)
In this analysis, the trend M(x) was assumed to be function of depth z, because the trend was
approximately invariable in the small analysis domain:
logK(x) = M1(z) + Re(x) xD
(5.9)
where D is the domain of the two-dimensional model, and M1(z) is the specific trend at the
observation well BW31. In-situ measurements of logK at BW31 were used as the honored data
119
in the conditional simulation; i.e., logK (m/s) = 2.777 (Z= 3.5 m), 2.955 (Z= 9.5 m), 4.249
(Z= 19.5 m), 2.668 (Z= 29.5 m), 5.359 (Z= 31.5 m), 3.754 (Z= 39.5 m), 5.521 (Z= 49.5 m),
and 4.320 (Z= 59.5 m), respectively.
120
a
Well Coordinates (m)
ID Y(E-W) X(N-S)
1 -73,307 -108,928
2 -72,005 -107,831
3 -74,506 -106,755
4 -72,260 -106,255
5 -71,196 -106,297
6 -74,877 -105,517
7 -71,939 -103,728
8 -70,835 -104,202
9 -72,351 -102,088
10 -73,463 -104,514
11 -73,637 -103,895
12 -72,012 -104,216
13 -72,603 -105,300
14 -72,734 -106,278
15 -72,993 -106,820
16 -73,005 -107,418
17 -73,164 -108,782
18 -72,296 -108,936
19 -71,420 -109,347
20 -72,996 -109,079
21 -72,846 -109,086
22 -72,019 -109,202
23 -72,019 -109,202
24 -71,119 -109,288
25 -73,262 -108,990
26 -72,150 -106,595
27 -69,988 -103,193
Test screen
Test result (m/s)
depth
logK logK 0
(m)
7.7 13.2 -3.55 -3.05
8.0 19.0 -6.37 -5.73
9.5 15.0 -3.14 -2.56
4.0 10.0 -5.70 -5.37
14.5 20.0 -4.53 -3.70
9.5 15.0 -2.62 -2.62
11.5 17.0 -2.60 -2.60
4.5 13.5 -2.65 -2.65
10.0 18.0 -3.66 -3.66
14.0 19.5 -2.97 -2.97
27.0 32.5 -3.06 -3.06
9.0 10.0 -2.82 -2.82
4.0 5.0 -4.28 -3.28
5.0 6.0 -2.24
8.0 9.0 -3.33
3.0 4.0 -2.78 -3.00
9.0 10.0 -2.95
19.0 20.0 -4.25
29.0 30.0 -2.67
31.0 32.0 -5.36
39.0 40.0 -3.75
49.0 50.0 -5.52
59.0 60.0 -4.32
63.0 64.0 -5.49
5.0 6.0 -3.39 -3.21
8.0 9.0 -3.69
3.0 4.0 -3.23 -3.78
9.0 10.0 -5.00
19.0 20.0 -3.85
29.0 30.0 -3.85
39.0 40.0 -4.80
49.0 50.0 -7.26
59.0 60.0 -4.63
69.0 70.0 -5.18
79.5 80.0 -6.82
89.0 90.0 -4.81
99.0 100.0 -5.13
6.0 7.0 -4.72 -3.34
9.0 10.0 -3.18
18.0 19.0 -3.78
8.0 8.5 -5.10 -4.64
20.0 20.5 -7.42
21.5 22.0 -3.79
10.0 10.5 -6.50 -4.28
21.0 21.5 -5.23
28.0 28.5 -3.97
8.0 8.5 -4.25 -4.21
8.0 8.5
14.0 14.5 -5.24
12.0 15.0 -5.72 -5.08
5.5 6.0 -2.41 -3.12
15.0 15.5 -4.83
13.0 13.5 -5.25 -4.32
21.5 22.5 -5.08
17.0 17.5 -4.74 -3.91
8.8 9.3 -4.75 -3.14
17.5 18.0 -3.20
26.0 26.5 -4.01
15.9 16.4 -5.56 -4.83
22.0 29.8 -5.53
34.4 35.0 -6.06
53.4 61.4 -5.63
58.7 59.2 -7.94
64.0 64.5 -7.98
76.3 80.8 -5.84
86.3 87.0 -4.92
97.8 99.2 -6.03
13.0 14.0 -4.90 -4.85
18.7 19.4 -4.98
32.5 33.2 -4.45
43.0 44.5 -4.84
73.0 76.0 -5.07
b
Well Coordinates (m)
ID Y(E-W) X(N-S)
28 -74,021 -107,143
29 -75,209 -103,982
30 -72,090 -110,835
31 -73,608 -110,881
32
33
34
35
36
-73,608
-110,881
-74,965
-110,681
-72,308
-106,653
-73,802
-105,518
-70,096
-102,729
37 -73,557 -102,370
38 -73,557 -102,370
39 -73,217 -102,634
40 -72,738 -102,965
41 -72,740 -103,181
42 -72,698 -103,490
43
44
45
46
-74,757
-103,530
-74,938
-103,559
-75,033
-103,897
-75,913
-105,708
48
49
49
49
50
51
-74,052
-101,299
-73,912
-101,378
-73,912
-101,378
-73,912
-101,378
-73,840
-100,880
-67,924
-103,080
52
53
54
55
56
57
58
59
60
-71,237
-103,516
-71,261
-103,411
-72,603
-104,564
-72,642
-104,411
-72,660
-104,300
-72,683
-104,182
-72,710
-104,019
-72,732
-103,895
-72,757
-103,773
61
62
63
64
65
66
67
-72,773
-103,646
-72,820
-103,429
-73,225
-103,339
-73,241
-103,535
-73,169
-103,599
-73,194
-103,762
-73,130
-103,900
47
Test screen
Test result (m/s)
depth
logK logK 0
(m)
14.9 15.9 -1.36 -3.70
35.5 36.0 -5.57
45.6 47.2 -4.72
54.7 55.2 -4.73
75.5 76.2 -8.26
78.3 78.8 -5.44
8.9 9.4 -4.40 -4.75
32.0 32.5 -4.19
51.9 52.4 -4.60
74.6 75.2 -5.64
91.0 92.6 -4.93
11.5 12.0 -4.82 -3.95
16.5 17.0 -4.45
8.5 9.0 -3.35 -3.29
16.0 17.0 -4.43
19.5 20.0 -5.49 -4.55
5.2 5.7 -5.07 -4.81
17.4 59.5 -4.20 -2.77
34.0 72.5 -3.47 -3.47
3.5 4.5 -4.51 -5.52
10.5 11.0 -6.52
2.5 3.5 -5.70
5.0 6.0 -6.94
2.5 6.0 -6.32 -6.32
7.0 7.5 -5.49 -5.08
11.0 11.5 -4.67
5.5 5.5 -5.03 -5.17
8.5 9.0 -5.31
5.5 5.6 -5.48 -4.84
8.5 9.0 -4.20
5.5 5.5 -4.42 -5.18
8.5 9.0 -5.94
10.0 11.0 -3.79 -3.79
10.0 11.0 -2.43 -2.43
10.0 11.0 -3.90 -3.90
7.0 7.4 -4.54 -5.29
11.0 11.5 -7.28
14.5 15.4 -5.48
17.8 18.6 -5.88
24.0 25.0 -3.17
28.1 29.0 -5.36
8.0 8.5 -3.86 -5.17
24.0 24.5 -6.49
8.0 24.5 -5.17
7.0 7.5 -5.60 -5.60
6.2 6.5 -5.02
8.4 8.9 -5.99
6.2 8.9 -5.50 -5.50
7.0 7.5 -5.38 -5.38
15.0 15.5 -2.71 -3.48
19.0 19.5 -3.90
26.5 27.0 -3.82
8.5 9.0 -4.49 -4.49
7.6 8.1 -5.13 -5.13
14.0 14.5 -4.09 -4.09
9.0 9.5 -4.96 -4.96
10.0 10.5 -3.06 -3.06
11.0 11.5 -5.10 -5.10
13.0 13.5 -5.01 -5.01
11.0 11.5 -5.25 -5.25
7.0 7.5 -4.65 -3.37
15.0 15.3 -2.10
7.0 7.5 -4.82 -4.82
9.0 9.5 -4.18 -4.18
9.5 10.0 -4.62 -4.62
9.5 10.0 -4.67 -4.67
10.5 11.0 -4.86 -4.86
10.5 11.0 -5.20 -5.20
11.0 11.5 -5.09 -5.09
121
5.2.4
A three-dimensional model is ideal for realizing actual groundwater flow and solute transport,
however, its computational cost is often effortful, especially in stochastic simulation. This
analysis applied a two-dimensional cross-sectional model. The simplified model provided no less
accurate solutions than the three- dimensional model in the typical condition that main lateral
direction of groundwater flow was only one direction. The GTE mound along the river showed
that the horizontal flows of groundwater were mainly perpendicular to the channel (Chapter 2).
The two-dimensional finite-difference models are shown in Figure 5.4. A SEAWAT program
was originally three-dimensional code, but this analysis utilized the program in the vertical
cross-sectional model, of which width is 1 m in all cells. The two-dimensional model was thus
the vertical cross sectional box, which was 190 m in length from the center of river and 64 m in
height (between 35 m asl and 29 m asl). The upper boundary was constantly 29 m asl because
the unsaturated zone was of insignificance in this study. The lower boundary was inclined
westward as estimated by the geologic information (Figure 5.7).
Two models of different discretization were prepared for a comparison in this study;
high-resolution model and block averaged model. The high-resolution model (Figure 5.4a) was
mainly used in this study. The model consisted of grid cells of two different sizes: 1 m square in
the lateral (x) and vertical (z) directions between the center of the river (x = 0 m) and the
observation well BW031 (x = 64 m); 5 m in the lateral direction by 1 m in the vertical direction
in the other area. The vertical sizes of cells in the bottom layer (below the elevation of 25 m)
were variable adjacent to the inclination of the basement. The K values in the cells were input
from the SGS realizations of 1 m regular grid, of which the hundredth place was rounded off. The
K values in the cells in the lowermost layer were geometric averages over the individual values in
each cell
The purpose of the fine discretization around the river was to satisfy the criterion of the Peclet
number for stable solution in the conduction-convection equation. Peclet number was generally
formulated in the case of heat transport as:
122
Pe
w c w qL
ke
qL
De
(5.10)
where w is the density of fluid, cw is the specific heat per unit volume of water, q is Darcys
specific discharge, L is a characteristic length, and ke is the effective (bulk) thermal conductivity.
Darcys specific discharge is calculated by multiplying K by hydraulic gradient. De is the
effective diffusivity. Considering one typical case; the hydraulic gradient was 0.1; K was 10-1 m/s
(i.e., Darcys flux q=10-2 m2/s); the grid-spacing L was 1 m; the bulk thermal diffusivity De
0.0083 m2/hr, the Peclet number was 1.2, and the value was less than the criterion value from 2
(Gray and Pinder 1976) to 4 (Anderson and Woessner 1992; Zheng and Bennett, 1995). The
Courant number was also utilized to determine the discretization in time, i.e., time step:
C= vdT/L
(5.11)
where C is the Courant number, v is the average linear velocity of groundwater, which is
obtained by the Darcys velocity dividing by the effective porosity, for example 0.2 in sandy
gravel. dT is the time step (1 hr in this study). If v is 10-2 or 10-3 m/s, the Courant number
obtained is 0.24 or 0.024, which are less than the criterion value, 1 (Anderson and Woessner
1992).
The block-averaged model (Figure 5.4b) was used in only two study cases for a comparison
with the high-resolution model. The grid size in the practical model was 5 m square throughout
the domain. The K values in the cells were geometric averages of the same values of
high-resolution model (i.e., SGS realizations) included in the cells.
A total calculation period for analysis was 264 days (i.e., 10 June 2010 to 28 February 2011).
The initial analysis date was the start date of the well observation (9 June 2010), and the final
analysis date was the end of last month when the temperature observation near the Horohira
Bridge stopped (21 March 2011).
123
River
BW03-2
BW03-4
BW03-1
W2
condition
BW03-5
W1
29
24
Elevation (m)
16
8
0
-8
-16
-24
-35
-140
-120
-90
BW03-2
BW03-4
-60
Distance (m)
-30
BW03-1
50
30
W1
BW03-5
29
24
Elevation (m)
16
8
0
-8
-16
-24
-35
-140
-120
-90
-60
-30
Distance (m)
30
50
124
Figure 5.4 Two dimensional models of high resolution (a) and of block average (b) in this study. Random colors in the cells indicate magnitudes of K
5.2.5
Boundary conditions along the left and right sides of the cross-sectional model were given as
the Dirichlet conditions. Under the condition, the hydraulic heads in the cells of the both sides
were consistently fixed at the specified heads. The groundwater tables in the both sides were
observed in BW34 and BW35. The vertical hydraulic gradient in the both sides were
calculated by using the discrepancies of heads between the deep BW31 and the shallow BW32.
No vertical fluxes into/from the external domain are assumed in the bottom of the model (no flux
condition). River boundary condition is a typical condition in the program. The condition aims to
simulate the influence of a surface water body on the groundwater flow. The MODFLOW River
Package input file requires the following information for each grid cell containing a River
boundary. The discharge from or to the river cells is calculated by Darcys law as:
(5.12)
where QRIV is the rate of leakage between the river and the aquifer: CRIV is the streambed
conductance to account the length L, and width W of the river channel in the cell, the thickness of
the bedding materials M, and their vertical hydraulic conductivity Kr,: h is the heads in the RBOT
is the bottom of the streambed. A schematic River condition is shown in Figure 5.5. The river
stages HRIV were not observed around the Horohira Bridge. This study observed the river stages
at the Horohira Bridge on 2010, and found the good relation with the stages at the upper Moiwa
gaging station (R2= 0.95). Therefore, the river stages for analysis periods were estimated from the
data at the upper station using the relation. The temperature of the river cells was those observed
at the Horohira-Bridge. As usual, it was difficult to determine vertical conductivity Kr in the
bedding materials due to its heterogeneity and widely variation in orders of magnitude
(Fleckenstein et al. 2006; Doppler et al. 2007; Genereux et al. 2008; Frei et al. 2009). For its
uncertainty, Kr was fixed at 100 m/d (110-3 m/s) in most study cases, and at 110-5 m/s except
for Case 3. The River cells were the upper cells between x= 10 m to 10 m (total width 20 m) in
most cases, and the cells between x= 20 to 20 m (40 m). The bottom of bedding material RBOT
was consistent at 28.5 m asl in all study cases.
125
Diffuse recharge flux were not given in other upper cells. The total precipitation during the
analysis period was 1,152 mm, and the total volume of water in the analysis domain was 218 m3.
On the other hand, the seepage loss in the distinct losing section (1.5 km) was determined at
about 1m3/s along the (Chapter 2), and so the loss in the model (unit length) during the period
was estimated at 1.510-4 m3. The seepage loss was over 70-folds of the precipitation, such that
the water volume of precipitation was enough small to be neglected in the focused recharge zone.
A list of input parameters used is summarized in Table 5.3. The storage coefficient was the
results from the previous pumping results (S=0.05), and the total porosity was determined by the
laboratory tests of the undistrurbed samples. The thermal properties (specific heat and thermal
conductivity) were the bulk values of andesitic gravel deposits in the literature (ICE 1976;
Jumikis 1977; Jessop 1990; Langevin et al. 2008). The bulk thermal diffusivity Dm was generally
anisotropic, and but the detail was not unknown. In this dissertation, the vertical diffusivity Dm
was one hundredth of the horizontal value.
126
Comments
12 only in case 8
Assigned
64
12 only in case 8
x= 1 m only in x= 0 to 64, except for case 8
x (DELR)
1 or 5
y (DELC)
Assigned
z (DZ)
Anisotropy ratio ( )
Specific storage (S s )
Aquifer properties
Units
190
Porosity ( )
m/d
1 or 0.1 or 0.01
m/d
1.0010-5
variable
the ratio of vertical hydraulic conductivity to horizontal value,
=1 in case1a and 2a, =0.01 in case1c and 2c, =0.1 in
0.20
Assigned
Longitudinal dispersivity ( L)
10
Assigned
Transverse dispersivity ( T)
0.1 or 10
835
J/(kg )
2,600
kg/m3
Bulk density ( b )
2,080
kg/m3
2.30
W/(m )
T= 10 m only in case5
1.96
W/(m )
0.202
m2/d
4,186
J/(kg )
0.58
W/(m )
Value
/T
Reference temperature (T 0)
Reference viscosity
/C
Reference concentration for viscosity (C 0)
2.0010
-4
m /kg
Assigned
0.7
kg/m3
Assigned
-0.375
kg/(m3 )
25
0.001
kg/(m s)
1.9210-6
m2/d
kg/m3
Assigned
1,000
kg/m
2.39410-5
A2
10
A3
248.37
A4
133.15
4.4610-3
kg/m4
/l
5.2.6
Study case
13 study cases were prepared for this analysis to assess the importance of modeling trending
heterogeneity and of determining other uncertain parameters (Table 5.4).
Case1 was the standard case of trending heterogeneity, of which realizations of K were
generated in the geostatistical methods as described in Chapter 5.2.3. Case1 consisted of three
sub-cases: Case1; vertical K (Kv) is equal to horizontal K (Kh) in all cells (isotropic condition):
Case1b; Kv is one tenth of Kh (anisotropic condition): Case1b; Kv is one hundredth of Kh (large
anisotropic condition); the anisotropy ratio was 1 (Case1a), 0.1 (Case1b), or 0.01(Case1c).
127
Case2 of stationary heterogeneity was prepared for a comparison with Case1. The realizations
of K in Case 2 were produced using the sane residuals Re(x) with Case1, but the trend component
M(x) was different as:
1
Z bottom
zbottom
M ( z )dz
1
Z bottom
logK(x)= M + Re(x) xD
(5.13)
(5.14)
Considering Zbottom= 64m, A= 0.11(m-1), and z0= 30m, the average trend M was 1.10 m/s.
The three sub-cases (Case2a, b, and c) were also prepared as well as Case1.
Case37 was performed to assess sensitivities of other uncertain factors as: vertical hydraulic
conductivity in the riverbed material, river width of seepage boundary, anisotropy of thermal
diffusivity, spatial variability of the residual component, and temperature-dependence of fluids.
The vertical hydraulic conductivity of the riverbed material, Kr, in Case3 was one-hundredth of
the standard value in Case1. The river width in Case4 was twice of the standard value in Case1.
The bulk thermal diffusivity in Case5 was isotropic. The lateral- and vertical- variogram ranges
of the residual in case6 were one tenth of the standard values in Case1. The density and viscosity
of fluid in Case7 were independent of temperature. Case8 also utilized block averaged model of 5
m square cells as shown in Figure 5.4b. Optimal realizations of K in each study case were
extracted among the 100 realizations through a comparison with calculation and measurements
results of groundwater temperature. Average root mean square error, RMSE , and average root
means square normalized error, RMSN , were used for error judgment. The individual RMSEj and
RMSNj in the observation points (index j) were calculated using calculated and observed
temperatures, T (t i ) and T ( ti ) as:
RMSE
1 6
RMSE j
6 1
(5.15)
RMSN
1 6
RMSN j
6 1
(5.16)
RMSE j
1 n
T j (t i ) T j (t i )
n i 1
(5.17)
128
RMSN j
RMSE
(max T j min T j )
(5.18)
n is the number of daily mean data (264 in this case), and j is the index of observation wells;
BW31 at a depth of 15 m (j=1); BW31 at a depth of 30 m (j=2); BW31 at a depth of 60 m
(j=3); BW32(j=4); BW34 (j=5); BW35 (j=6). The max Tj and min Tj are the maximum and
minimum of Tj, respectively. The good realizations of K in each study case were extracted
when RMSE and RMSN were less than 2 and 1, respectively.
129
a
Trending heterogeneity
b
Anisotropy
K r in riverbed
River width
Case5
Case6
Case7
Case8
a
b
Temp. dependence g
Block averaged
Case3 Case4
Thermal diffusivity e
Variogram
Case2
b
c
trending heterogeneity, b anisotropy ratio is 0.1 and anisotropy ratio is 0.01, c vertical hydraulic conductivity of bedding material is one tenth of standard value
(10-5 m/s at 25 C), d river width is twice of standard value (20 m) , e thermal diffusivity is isotropic, f lateral- and vertical- variogram range of residuals are one tenth of
standard value, g density and viscosity of fluid are independent of temperature, h blocked averaged model of 5 m square cells is used instead of high-resolution model
130
5.3
5.3.1
131
30
Horohira
BW3-2
BW3-4
BW3-5
BW3-1
BW3-2
BW3-4
BW3-5
25
BW03-1
20
Jun
2010
Jul
Aug
Sep
Oct
Nov
Dec
Jan
2011
Temperature
20
0
Jun
2010
BW3-1(60m)
BW3-5
BW3-2
Jul
Mar
BW3-1(15m)
BW3-1(30m)
BW3-1(60m)
BW3-2
BW3-4
BW3-5
River
River
10
Feb
BW3-1(30m)
BW3-1(15m)
BW3-4
Aug
Sep
Oct
Nov
Dec
Jan
2011
Feb
Mar
Figure 5.6 Hourly water-level and temperature variations in the river and observation wells during the analysis periods; 10 June 2010 28 February
2011 (264 days)
132
Figure 5.7 Contour maps of water level and temperature in the vertical cross section
133
9 June 2010
9 Aug. 2010
10
2 Dec. 2010
Depth (m)
20
30
40
50
60
5
10
Temperature ()
15
134
5.3.2
Figure 5.9 shows relations between hydraulic conductivity at the groundsurface, logK0, and
coordinates in JGD2000, X (NS direction) or Y (EW direction) coordinates. There was no
obvious auto-correlation in the EW direction as shown in Figure 5.9a. The coefficient b in Eq.
(5.3) was approximately zero to be neglected throughout the fan. This also indicated the validity
of simplification in this analysis; the lateral trend in the two-dimensional model was assumed to
be constant because the vertical cross-sectional model was amost parallel to the Y (EW)
direction.
Figure 5.9b showed a certain correlation between the hydraulic conductivity and the coordinate.
The coordinate was correspondent to the downward direction of the terrain, but the correlation
was not uniform, and the highest values of K were obtained at the middle of the fan. Figure 5.9b
probably indicated the humped trends of K as in other fans. The term humped trend means that
the maximum K appears not in the fan-apex, but in the middle fan due to the sorting effect in
depositional process. The typical hydrogeologic structure was frequently seen in alluvial fans
around the world (Neton et al. 1994). The coefficient a and c of trend component Mh(x) in Eq.
(5.3) were determined by the linear regression analysis; a= 0.00303 (1/s) and c= 29.2 (m/s) at X<
106,000 m or a= 0.00583 1/s and c= 64.7 m/s at X> 106,000 m. In this analysis, the lateral
trend in the two-dimensional model, M1(z), was constant at 3.07, that was at BW31 (X=
106,278 m). The expectation of K at the groundsurface was stationary at 10-3.07 m/s in the
two-dimensional model, indicating extremely high permeability in the fan-deposits near the
surface.
135
logK0 (m/s)
-2
-4
-6
-75,000
-70,000
logK0 (m/s)
-2
X=-106,000
logK0=29.2+0.000303X
-4
-6
logK0=-64.7-0.000583X
-110,000
-105,000
-100,000
Figure 5.9 Relation between hydraulic conductivity at the ground surface, logK0,
and coordinates in JGD2000; a) Y (EW) and b) X(NS) coordinates
136
5.3.3
Variogram analysis
The experimental and theoretical variograms of residual are shown in Figure 5.10. The model
parameters in the lateral model (a) were a= 1000 m, B= 1.15 m/s, and C= 0.5 m/s, while the
parameters in the vertical model (b) were a= 20 m, B= 0.92 m/s, and C=0.5 m/s. The range a in
the horizontal direction was 50-folds of that in the vertical direction. The large anisotropy was
found in alluvial coarse deposits such as braided river system as shown in Deutsch (2007).
Variogram
1.5
0.5
500
1000
1500
2000
Lag Distance
137
5.3.4
The calculated values of water level and temperature in the 100 realizations of logK in 13
study cases are given in Appendix II. The optimal results are indicated in color when the results
satisfy the criterions of both RMSE and RMSN . Appendix III shows examples of optimal
realizations in each study case at the last period (ti= 264 day). All the results in Appendix III
commonly showed that infiltrated water from the river cell widely moved toward the both sides
of boundary. The directions of groundwater flows were variable, depending on the large
heterogeneity, while the magnitudes were accumulated in the shallow zone because of the
trending heterogeneity. As a result, the groundwater temperature contour includes sharp fronts
reflected to the preferential flows, especially in the fine-spacing area, i.e., x= 63 to 0 m in
distance. The block averaged models in Case8 indicates more smooth and uniform contours in
hydraulic head and groundwater temperature.
(1) Comparison between trending heterogeneity, stationary heterogeneity and
block averaged heterogeneity
Figure 5.11 shows simulation results of two realizations of logK (no.11 and no.33) in Case1
(trending heterogeneity), Case2 (stationary heterogeneity), and Case8 (block averaged
heterogeneity). Here, no.11a (upper-left) and no.33b (lower-right) were the optimum realizations
in Case1a (isotropy) and Case2 (anisotropy), respectively. The high-resolution images in Case1
and Case2 indicated large heterogeneity, reaching low values (logK< 9) near the bottom and
high values (logK> 1) near the ground surface. The horizontal zones of low values between 15
m and 20 m in depth were less than the honored data of low logK (logK= 4.249 m/s at a depth
of 19.5 m and a distance of 62 m). The optimal realization was just one case in the stochastic
simulation that required various assumptions and limitations, however, the realizations could
provide the optimum calculations of groundwater temperature, probably due to preferential flow
zone above the low-permeable thin layers
The realizations in the upper and the middle images were well-agreed below a depth of 30m,
which was the maximum depth of trending heterogeneity, i.e., z0. In the shallower depth, on the
other hand, the middle realizations had a little narrower range of logK. This is because of no
trending stationary in Case2. The low value of logK in the shallow zone probably prevented to
138
139
Realization no.16
Realization no.33
log10K
20
Depth (m)
Depth (m)
20
40
0
-1
-2
40
-3
-4
60
-5
60
-6
-120
-80
-40
Distance (m)
40
-120
-80
-40
Distance (m)
40
-7
-8
-9
-10
20
Depth (m)
Depth (m)
20
40
60
60
-120
-80
-40
Distance (m)
40
-120
-80
-40
Distance (m)
40
20
Depth (m)
20
Depth (m)
40
40
40
60
60
-120
-80
-40
Distance (m)
40
-120
-80
-40
Distance (m)
40
Figure 5.11 Examples of two-dimensional stochastic realizations of logK: the upper images (a) indicate trending heterogeneity in Case1; the
middles (b) indicate stationary heterogeneity in Case2, and the lowers (c) indicate block averaged heterogeneity in Case8. The left
and right images are the optimal realizations, no.11 in Case1a and no.33 in Case1b, respectively, where no. denotes individual
number among 100 realizations
140
Table 5.5 Individual (middle lines) and total numbers (bottom line) of optimum
realizations in each study case (upper line)
1a
1b
1c
2a
2b
2c
16
13
63
47
47
8a
8b
63
70
78
92
17
18
25
30
33
42
44
63
65
78
14
141
Hydraulic head
The well-agreements between calculations and measurements in the deep well (BW31) and
the shallow well (BW32) were seen in the cases of trending heterogeneity, Case1a and Case1c
(RMSE< 1m). The results in the case of Case1b are inferior to the agreement, and the RMSE
values were about 2 m.
2)
River leakage
Assuming that the river leakage was uniform in space and constant during the analysis period,
the leakage value in unit length (model width) was estimated at 57.6m3/d. The calculation in
Case1a was highest and closest to the estimate. The calculation in Case1b and Case1c were about
one-tenth of the estimate, and so these values might be too small. However, the uniform and
constant leakage was based on the excessive assumption, and the validity of calculated leakage
was determined in this study.
3)
Groundwater temperature
The calculations in three different depths at BW31 were as much stable as the observations in
all of the cases. Also in BW35, the calculations were sufficiently equal to the measurements,
although the daily fluctuations measured were smooth in the calculations. In BW32, the
calculations in Case1a and Case1b were matched with the measurements, but the calculation in
Case1c was largely different from the measurement. Especially in BW34, only the calculation
in Case1a obtained the sufficient agreement with the measurement (RMSE= 0.3 C).
4)
Summary
Compared with the optimal realizations in the isotropy, anisotropy, and large anisotropy
conditions, the isotropic realization in Case1a was most reasonable in terms with agreement with
measurements of both hydraulic head and groundwater temperature. The river leakage in the
142
isotropic condition was several-folds smaller than the estimate in Case1a. However, the
discrepancy was of less importance because the seepage loss was determined only through the
distinct losing section, and the detailed value in the local area was not unknown in this study.
Figure 5.11 shows the low-permeability layer was interbedded between 15 and 20 m in depth the
high-permeable gravel deposits. Information about the interbedding fine layer, instead of
anisotropy in cells, was probably required for obtaining plausible realizations. The example
results calculated in the isotropic condition were shown in Appendix II. The groundwater head
contours were densely distributed in the thin layer, and the sharp fronts of water temperature
accumulated in the shallow zone above about 15 m in depth. On the other hand, both of
anisotropic cases also indicated insufficient agreements with the measurements, especially of the
water temperature in the shallow wells. Consequently, information about both of trending
heterogeneity and hydrogeological structure is required to obtain the plausible descriptions of
complex groundwater flow and heat transport system.
143
30
Obs._River
Case1a
RMSE=0.5m
25
Case8a
RMSE=0.5m
Case8b
RMSE=2.1m
Case1b
RMSE=2.0m
Case8b
RMSE=4.3m
Case8a
RMSE=1.9m
Case1b
RMSE=2.5m
25
25
10
Case8b
200
250
0
20
Heat (BW3-1:Z=15m)
Case8b
Case1b
RMSE=0.5 RMSE=0.3
15
Temperature ()
Case1a
RMSE=0.8
10
10
Obs._well
Case8a
RMSE=0.9
20
50
100
day 150
200
250
Heat (BW3-2:Z=5m)
Case1a
RMSE=0.5
Obs._well
15
Temperature ()
10
10
Case8b
RMSE=3.3
Case8a
RMSE=2.7
0
50
100
day 150
10
250
20
15
Case1a
RMSE=0.9
Case8b
RMSE=0.5
10
Case1b
RMSE=1.8
200
10
Obs._well
Case8a
RMSE=0.7
20
50
100
20
day 150
200
250
Case1a
RMSE=0.3
Obs._well
150
200
250
10
20
Obs._River
15
Case1a
RMSE=0.2
10
10
Case1a
RMSE=0.2
10
Case8b
RMSE=3.4
day
Obs._well
15
100
Heat (BW3-1:Z=60m)
Case8b
RMSE=0.1
Case1b
RMSE=0.2
5
20
50
100
day 150
200
250
0
20
Heat (BW3-5:Z=4m)
Obs._River
15
50
15
20
Heat (BW3-4:Z=10.5m)
10
Obs._River
Case1b
RMSE=0.6
Obs._River
15
Case8a
day 150
15
20
100
Heat (BW3-1:Z=30m)
Obs._River
15
20
50
Temperature ()
day 150
10
20
Obs._well
Obs._River
Case1a
RMSE=1.3
15
Case1b
RMSE=1.4
15
Obs._well
Temperature ()
100
Temperature ()
20
50
Case1a
RMSE=0.7m
20
Temperature ()
Case1a
1
Case1b
20
20
10
3
Obs. 57.6m /d
Head (TPm)
Head (TPm)
Obs._well
30
Recharge (m /day)
30
Obs._River
25
10
Case8a
RMSE=4.3
10
10
Case8b
RMSE=5.6
Case1a
Case1b
RMSE=3.7 RMSE=3.6
200
250
50
100
day 150
200
250
50
100
day 150
200
250
Figure 5.12 Calculations vs measurements in hydraulic head, river leakage, and groundwater temperature, in the optimum realizations no.16 in
Case1a, no.4 in Case1b, and no.13 in Case1c
144
145
Obs._River
Case1a
RMSE=0.5m
25
Case8a
RMSE=0.5m
Case8b
RMSE=2.1m
Case1b
RMSE=2.0m
Case8b
RMSE=4.3m
Case8a
RMSE=1.9m
Case1b
RMSE=2.5m
25
25
10
Case8b
200
250
0
20
Heat (BW3-1:Z=15m)
Case8b
Case1b
RMSE=0.5 RMSE=0.3
15
Temperature ()
Case1a
RMSE=0.8
10
10
Obs._well
Case8a
RMSE=0.9
20
50
100
Heat (BW3-2:Z=5m)
day 150
200
250
Case1a
RMSE=0.5
Obs._well
20
Temperature ()
Case8b
RMSE=3.3
Case8a
RMSE=2.7
0
50
100
day 150
Case8b
RMSE=0.5
Obs._well
Case8a
RMSE=0.7
20
Case1a
RMSE=0.9
10
10
20
10
10
10
250
15
Case1b
RMSE=1.8
200
50
100
20
day 150
200
250
Case1a
RMSE=0.3
Obs._well
10
Case8b
RMSE=3.4
day
150
200
250
10
20
Obs._River
15
Case8a
RMSE=0.2
Obs._well
10
10
Case1a
RMSE=0.2
Case8b
RMSE=0.1
Case1b
RMSE=0.2
15
100
Heat (BW3-1:Z=60m)
20
50
100
day 150
200
250
0
20
Heat (BW3-5:Z=4m)
Obs._River
15
50
15
20
Heat (BW3-4:Z=10.5m)
10
Obs._River
Case1b
RMSE=0.6
15
Case8a
day 150
15
Obs._River
15
100
Heat (BW3-1:Z=30m)
Obs._River
15
20
50
Temperature ()
day 150
10
20
Obs._well
Obs._River
Case1a
RMSE=1.3
15
Obs._well
Temperature ()
100
Temperature ()
20
50
Temperature ()
Case1a
RMSE=0.7m
20
20
Case1a
Case1b
20
10
3
Obs. 57.6m /d
Head (TPm)
Head (TPm)
Obs._well
30
Recharge (m /day)
30
30
Obs._River
25
10
Case1b
RMSE=1.4
Case8a
RMSE=4.3
10
15
10
Case8b
RMSE=5.6
Case8a
Case1b
RMSE=3.7 RMSE=3.6
200
250
50
100
day 150
200
250
50
100
day 150
200
250
Figure 5.13 Calculations vs measurements between original and block averaged models using the same realizations of hydraulic conductivity;
no.16 in Case1a and no.4 in Case1b
146
5.4
147
different locations and depths. Several optimal realizations of K were obtained in the
nonstationary condition, but no optimal realization was obtained in the stationary condition, even
though the same 100 realizations of residuals were used in both conditions. This tendency was
the same whether the model included isotropy, anisotropy, or large anisotropy. These simulation
results indicate that modeling of trending heterogeneity was indispensable to obtain plausible
descriptions of groundwater flows and solute transport in the focused recharge zone. In particular,
in a comparison between isotropic, anisotropic, and highly anisotropic conditions under an
assumption of nonstationary, the most plausible realizations of K were obtained in the isotropic
condition. Compared with other case studies of different conditions in the high-resolution model,
the numbers of optimal realizations obtained were nearly the same as in cases under the
nonstationary condition. This indicated that, compared with other factors, heterogeneity, that is
whether there was trending heterogeneity or stationary heterogeneity, was more significant for
modeling the groundwater flow system in the focused recharge zone.
The calculation of hydraulic head, river leakage, and groundwater temperature was compared
in the optimal realizations of isotropy, anisotropy, and large anisotropy cases. Only the isotropy
condition yielded the reasonable calculations of hydraulic head and groundwater temperature,
and other anisotropic conditions resulted in the unacceptable discrepancy of temperature in the
far- and shallow- well. The most plausible realization in isotropy condition suggested the
low-permeable thin layer might be interebdded in the depth of 1520m. The detailed information
about hydro-facies is also important as well as trending heterogeneity in the gravel deposits.
The block-averaged models consequently provided smoother and more uniform descriptions of
the groundwater flow and heat-transport system, indicating no preferential flows. However, the
numbers of optimal realizations obtained were larger than those obtained in the high-resolution
model. The superiority of the block-averaged model indicated that conventional geometric
averaging was similarly effective for upscaling the model of the heterogeneous alluvial gravel
deposits. Upscaling such as block averaging was needed to construct more regional model in the
limitation of computational power. The upscaling process has commonly faced the groundwater
studies, and the last chapter discussed the future direction to address the problem. The anisotropic
condition in the block cells was a typical important factor, but the anisotropy ratio was constant
148
in this analysis. In fact, the ratio was different from cell to cell as a result of heterogeneity. In
other words, the upscaling procedure must be understood as a tensor. If this can be done, more
plausible realization can be expected, and this will be pursued in future work.
This chapter applied two-dimensional vertical cross sectional model in the assumption that the
major horizontal components of groundwater flow vector were perpendicular to the river channel.
In more rigorous term, orthogonal flows from the upstream to the downstream affects on the
groundwater flow and transport system. Alternative three-dimensional groundwater flow model is
needed to discuss the importance of trending heterogeneity in more detail.
149
Conclusions
Water is indispensable to living and its value is increasing, especially in arid and semi-arid regions.
Among various geologic settings, alluvial coarse deposits form abundant reservoirs of groundwater
around the world. However, the capacity is exhaustible because of area limitations. Furthermore,
such reservoirs are susceptible to contamination and dewatering due to their high permeability.
Alluvial coarse deposits consist of poorly sorted gravel deposits, and are bounded by steep land
slope surfaces. Hence, the geologic heterogeneity is typically large and trending. Geostatistical
approaches have been widely used as a powerful method for addressing uncertainty in the
heterogeneity of various geologic settings. However, trending heterogeneity cannot be addressed by
conventional geostatistical methods, which require an assumption of stationarity. Consequently,
geostatistical reservoir modeling of alluvial coarse deposits has proved especially difficulty. A
standard approach to nonstationary problems is to separate the attribute into a global trend
component and a residual component, but the global trend has rarely been determined because of
insufficient measurements of the attribute, especially in the deep parts of the reservoir.
The study area covered in this dissertation was a focused recharge zone beneath and around a
losing river. In the focused recharge zone, water that infiltrated from the losing river had primarily
vertical flow components, and the flow vectors varied in space and time, depending on the trending
heterogeneity. If the zone has trending heterogeneity, for example, a coarsening-upward trend, the
groundwater flows will accumulate in the shallower zone, and then groundwater temperature will be
indicated by typical contours reflected directly from the trending heterogeneity. The representative
zone was just a portion of the regional system of groundwater flows, but detailed research in the
zone enabled the importance of trending heterogeneity in the groundwater flow system to be
assessed by restricting various uncertain factors in groundwater models to only trending
heterogeneity of the deposits.
Targeting the Toyohira River alluvial fan (Sapporo, Hokkaido, Japan), this dissertation presented
well data analysis, flow measurements, permeability modeling, and stochastic simulations. First,
vertical hydraulic gradient was mapped based on groundwater table elevation throughout the fan.
Uncertainty of water-level data in individual wells and boreholes was addressed in this study by the
150
following means: data filtering based on long-term water-level variations, kriging interpolation
assuming nonstationary conditions, and cross-validation.
A synoptic survey of the river flow was also performed to determine the magnitude and
distribution of seepage loss from the Toyohira River. Synoptic surveys must usually contend with
uncertainty in discharge measurements, especially in gravel-bed rivers with unsteady turbulent flow
and rough boundaries. In this work, the uncertainty was reduced as much as possible by combining
a handheld acoustic flow velocimeter and detailed arrangements of verticals for measurements of
flow and depth. As a result of the improvements, the relative uncertainty of discharge measurements
was only 24%.
This dissertation also has established the permeability model of the alluvial fan gravel deposits
through the use of relatively undisturbed cores collected from the fan. The packing of fine sediments
between gravel grains was qualitatively indicated by a newly defined index, the matrix packing level.
The permeability model admitted a well-defined relation between slug test results and core
properties at a given depth. Then, the profiles of core properties at three locations that differed in
topographythe apex (BW5), the mid-fan (BW3), and the distal part (BW7)were translated into
profiles of hydraulic conductivity. The moving average method was also performed to reduce the
uncertainty in the estimated values. Consequently, a common depth dependence of the permeability
in the apex and the mid-fan was formulated as an exponential model of hydraulic conductivity.
Finally, stochastic simulation was performed to assess the requirements for modeling trending
heterogeneity in order to describe the complexity of groundwater flows and solute transport in the
focused recharge zone. To obtain more rigorous solutions, the computer program SEAWAT was
used, in which fluid properties (density and viscosity) were dependent on groundwater temperature.
For the data preparation, the water levels and the temperatures were measured with piezometers
along the off-stream transect in the middle of the distinct losing section. Next, the horizontal trend in
hydraulic conductivity was analyzed by using previously conducted in situ measurements, and
variogram analysis of residuals was also performed. Sequential Gaussian simulation then produced
100 realizations of hydraulic conductivity by using the vertical trend and the variograms of residuals.
This study adopted 13 study cases, respectively, of different conditions such as trending or stationary
heterogeneity, isotropy or anisotropy of hydraulic conductivity, river conditions (hydraulic
151
conductivity and width), range of the variogram, anisotropy of thermal dispersivity, temperature
dependence or independence of fluid properties, and model discretization.
In summary, this thesis has revealed the hydrologic and hydrogeologic characteristics of the
groundwater flow system of the Toyohira River alluvial fan, particular around the focused recharge
zone. The finds are summarized as follows.
1) The resulting vertical hydraulic gradient and groundwater table elevation maps indicated the
hydrology of the fan featured downward groundwater flows, probably due to the effects of
urbanization. The negative peak of vertical hydraulic gradient along the losing river in the mid-fan
was associated with both the focused recharge and the dewatering by the subway lines crossing
below the zone.
2) The distinct losing section was found to be located in the section KP 15.517.0 (about 1.5 km
long), and the rate of water exchange was estimated at about 1 m3/s, which corresponded to about
80% of the total pumping of groundwater throughout Sapporo City. This means that the focused
recharge plays an important role in maintaining the groundwater reservoir.
3) The permeability model of the alluvial fan gravel deposits was established as a function of
effective grain size (d20) and length fraction of low packing levels in relatively undisturbed cores.
The profiles of estimated hydraulic conductivity K indicated clear decreasing trends in the fan
deposits with depth at the apex and the mid-fan, while no vertical trend was indicated in the distal
part.
4) The decreasing trend of permeability was formulated as an exponential model. The exponential
decay constant of 0.11 m-1 was 10- to 1000-fold of that of consolidated rocks reported in the
literature. The primary reason for the large value was depositional processes such as progradation.
Furthermore, post-depositional processes such as fine sediment infiltration might be a secondary
reason.
5) The longitudinal cross section of the Toyohira River alluvial fan was generated by using profiles
of moving average YMA. The moving average boundaries of YMA = 4 and YMA = 3 revealed the
hydrogeologic structures in the monotonic gravel sequences: an aquifer with a high infiltration rate
(Holocene no. IIa); a depth-dependent heterogeneous aquifer (Holocene no. IIb); and a stationary
permeable aquifer (Pleistocene no. III).
152
6) There were wide variations in hydraulic head and groundwater temperature near the losing
river, depending on the location and depth of the observation well. The seasonal envelope in the
groundwater temperature profiles was above a depth of about 30 m. These observations indicated
that the downward flows of groundwater were largely attributable to the focused recharge zone.
7) The horizontal trend in hydraulic conductivity in the fan showed a humped trend as has been
observed in other alluvial fans. Variogram analysis also revealed large anisotropy of spatial
variability, and the theoretical models fitted were spherical models with a range of 1,000 m in the
lateral direction and 20 m in the vertical direction.
8) The high-resolution model effectively realized various complex distributions of groundwater
flow vectors and temperature contours. The calculated heads and temperatures based on the
plausible realization satisfactorily corresponded to the measurements in the observation wells at
different locations and depths. Several optimal realizations of K were obtained in the nonstationary
condition, but no optimal realization was obtained in the stationary condition, even though the same
100 realizations of residuals were used in both conditions. This tendency was the same whether the
model included isotropy, anisotropy, or large anisotropy. These simulation results indicate that
modeling of trending heterogeneity was indispensable to obtain plausible descriptions of
groundwater flows and solute transport in the focused recharge zone. In particular, in a comparison
between isotropic, anisotropic, and highly anisotropic conditions under an assumption of
nonstationary, the most plausible realizations of K were obtained in the isotropic condition.
9) Compared with other case studies of different conditions in the high-resolution model, the
numbers of optimal realizations obtained were nearly the same as in cases under the nonstationary
condition. This indicated that, compared with other factors, heterogeneity, that is whether there was
trending heterogeneity or stationary heterogeneity, was more significant for modeling the
groundwater flow system in the focused recharge zone.
10) The block-averaged models consequently provided smoother and more uniform descriptions
of the groundwater flow and heat-transport system, but the numbers of optimal realizations obtained
were larger than those obtained in the high-resolution model. The superiority of the block-averaged
model indicated that conventional geometric averaging was similarly effective for upscaling the
model of the heterogeneous alluvial gravel deposits. The anisotropic condition in the block cells was
153
a typical important factor, but the anisotropy ratio was constant in this analysis. In fact, the ratio was
different from cell to cell as a result of heterogeneity. In other words, the upscaling procedure must
be understood as a tensor. If this can be done, more plausible realization can be expected, and this
will be pursued in future work.
6.2
The geostatistical reservoir model gave plausible realizations of the complex system of
groundwater flow and heat transport in the focused recharge zone of the Toyohira River alluvial fan.
However, an assumption of the vertical cross-sectional two-dimensional model was that the
orthogonal flows from upstream to downstream were negligible. However, this assumption is not
necessarily valid in alluvial coarse deposits because of the steep terrain in the longitudinal direction.
Thus, a next study is to extend the two-dimensional cross-sectional model to a three-dimensional
local model around the focused recharge zone, and to a three-dimensional basin model of the entire
fan. Figure 6.1 shows a schematic three-dimensional basin model of the alluvial fan. The
geostatistical process presented in this thesis also enables stochastic realizations of K to be produced
throughout the fan; however, plausible realizations cannot be obtained if a number of realizations are
generated without incorporating uncertain modeling factors that were assumed to be negligible in
the focused recharge zone. Lastly, some research directions are discussed in terms of several
modeling factors; hydrogeologic sequences, diffuse recharge, urbanization effects, and upscaling
procedure.
Humped trend in the longitudinal direction
Exponentially decreasing
trend in the vertical direction
154
155
156
artificial constructions and activities throughout the region is rarely available, and so large
uncertainty is commonly encountered in groundwater modeling of urbanized areas. In the Toyohira
River alluvial fan, there are two uncertain factors regarding urbanization: the subway lines and the
pumping wells. The monthly subway drainage amounts are recorded at each station by Sapporo City,
and the subway lines can be modeled as tunnel models by using the records. Modeling of numerous
pumping wells is more problematic because only a portion of individual wells are characterized in
terms of location, well depth, and pumping rate. Stochastic simulation of the well location and
pumping program might be effective for assessing urbanization dewatering.
157
Figure 6.2 Schematic diagram illustrating the scale-up procedure and the
upscaling analysis (Zhang et al. 2006)
158
Appendix A
Explanatory theory, equations, and knowledge utilized in this dissertation
A.1 Governing equation of groundwater flow
A.2 Gravitational flow system of groundwater in an alluvial fan
A.3 Facies model in coarse alluvial deposits
A.4 Spatial trend of hydrogeology in an alluvial fan
A.5 Interaction between surface water and groundwater
A.6 A plot method for detection of recharge/discharge zone
A.7 Mann-Kendall test and Sens slope estimation
A.8 Geostatistics for groundwater modeling
A.9 Theory of heat transport in groundwater
159
A.1
Ground water flow occurs because of the difference in energy of the water from one point to another.
The flow energy of water at a given point consists of potential energy, elastic energy and kinetic energy.
In general, the kinetic energy can be negligible because the ground water flow velocity is very low. The
convenient way to measure the energy of the water is to measure the ground water level or hydraulic head
in a piezometer (a cased well opened only over a very short section of its length). The level to which water
rises in the piezometer with reference to a datum such as sea level is the hydraulic head. The hydraulic
head is mathematically expressed as a sum of elevation head (potential energy per unit weight of water)
and pressure head (elastic energy) :
h=z+
(A.1.1)
where h is the hydraulic head [L], z is the elevation head [L], is the pressure head [L]. The difference
in hydraulic head between two points implies a hydraulic gradient. The gradient is along the direction of
the lower head. The horizontal component of the hydraulic gradient in a given area can be measured by
installing three or more piezometers to the same level in the ground, measuring the hydraulic head in each
piezometer, and contouring these head values. It follows that the gradient is in the direction perpendicular
to the hydraulic head, or equipotential contours, toward the decreasing hydraulic head. The fluid flux q
[LT-1] (often referred to as the Darcy flux) is given by Darcy's law:
Kx
qKh Kyx
Kzx
g
kr
K
Kxy
Ky
Kzy
dx
Kxz
dh
Kyz
dy
Kz
dh
dh
dz
(A.1.2)
(A.1.3)
kkr
q
pg
(A.1.4)
The formulation of the general groundwater flow equation in porous media is based on mass
conservation and Darcy's law, and can be represented in three dimensions as follows (e.g. Bear 1979;
Mercer and Faust 1981; de Marsily 1986):
h h h
h
(A.1.5)
Kx Kx Kx R S S
x x y y z z
t
160
where Kx, Ky and Kz [LT-1] are the hydraulic conductivities (or permeability) along the x, y and z
coordinates, respectively, assumed to be the principal directions of anisotropy; h [L] is the hydraulic head
given by h=p/g + z, that is to say, a sum of pressure head and elevation head, p is pressure, is the
density of water and g is gravitational acceleration at an elevation z (i.e. the water level measured in
piezometers; R' [T-1] is an external volumetric flux per unit volume entering or exiting the system; and SS
[L-1] is the specific storage coefficient.
The governing equation of groundwater flow indicates that conceptualization of groundwater
flow system corresponds to determination of variability of hydraulic properties and recharge flux in
space and time. The hydraulic properties, even though only in saturated zone, consist of hydraulic
conductivity, specific storage, water contents, total and effective porosity, and others. The hydraulic
conductivity is definitely most important because its magnitude proportionally affects on Darcys
flux, and the variation ranged in over 14 orders of magnitude due to various factors in geologic
material (Figure A. 1). In addition, Hydraulic conductivity depends on scale, that is, microscale (e.g.,
drilling cores; several meters or less), mesoscale (e.g., facies; 10 to 100 m), and macroscale (basin
scale, 100 m or more) as shown in Figure A. 2. The discrepancy between the different discretized
models shows a direction of upscaling in groundwater modeling of alluvial coarse deposits. Thus,
uncertainty in spatial variability of hydraulic conductivity faces groundwater modeling study.
161
162
A.2
Hydraulic head is represented as a sum of two components: elevation head and pressure head as
described in previous section. Infiltrated water in high land at precipitation has relatively high value
of elevation head, and groundwater flows laterally to lower land and vertically to deeper zone. Thus,
groundwater flow system is actually three dimensional. Hubbert (1940) was the first to present a
graphics of a typical flow net. Tth (2009) mathematically represented groundwater flow system by
analytical solutions of boundary-value problems using Laplaces equation. Tth (2009) also divided
groundwater flow system into local, intermediate, and regional flow systems, as shown in Figure A.
3. Groundwater flow becomes more complex as the topography and geology become more complex.
Local topography creates local ground water flow systems. The groundwater at a greater depth still
eventually reaches the major valley through the regional groundwater system. Also, geologic structure
may affect on the groundwater flow patterns. The groundwater tends to flow along high-permeable
zones and across low-permeable zone.
163
The mechanism of groundwater flow from apex to distal part in an alluvial fan is firstly described
in Tolman (1937). An alluvial fan is hydrologically divided into four areas; I) intake area; II) Upper
conduit zone; III) Lower conduit or Pressure zone; IV) Discharge zone. Intake area is shown as a belt
covering the upper, where influent stream seepage is usually the chief source of supply of unconfined
and confined groundwater. Upper conduit zone starts at the upper edge of a confining formation, and
furnishes a less satisfactory water supply than the downstream zone due to the long distance of water
table below the surface. Lower conduit or pressure zone is the zone in which the water in wells originally
rose to or above the surface. Discharge Zone is the leaking portion of the pressure zone, and may play out
entirely due to the constituent materials may gradually increase in fineness. Confined water moves down
the conduits and is discharged by leakage to the surface throughout the entire discharge zone. On the
other hand, free water moves down the water-table slope to the point where it intersects the surface, and
is discharged at the upper edge of the swamp or spring.
164
A.3
Facies model is commonly used to analysis sedimentary structure in various consolidated and
unconsolidated sediments. Monotonic and no-migration sequences in alluvial coarse deposits can be
subdivided as shown in Table A. 1 (Rust 1979). Among various facies, clast-supported and
matrix-supported gravel are of typical importance in coarse alluvial deposits (Figure A. 5). Clast
supported means the clasts in the sediments, i.e. pebbles, rocks, boulders etc., are touching each
other, indicating energetic aqueous transport while sand is still carried in suspension; as flow
velocity decreases, sand infiltrates into the spaces between the larger particles. Among the
depositional processes, sieve deposits form as lobes of the fans which receive little sand or mud
from their source areas. Open framework gravel (OFG) added into facies model is paid special
attention in various studies of sedimentary and hydrology studies (Jussel et al. 1994; Heinz 2003; Lunt
et al. 2004; Zappa et al. 2006; Lunt and Bridge 2007; Ferreira et al. 2010; dellArciprete et al. 2012). OFG
includes connective and open voids between gravels, such that its hydraulic conductivity is
extremely high, although OFG has a distribution that is only centimeters or decimeters thick.
Matrix-supported gravels are subdivided into two types: stratified sand matrix, and unstratified
silty sand matrix. The former indicates aqueous lower transport, but at a low energy level in which
sand and finer gravel particles are deposited together (Gp). The latter type of matrix support points
to deposition from massive flows, which in the alluvial context are debris flow (Gms).
Matrix-supported Gravel has relatively low permeability because the open spaces next to large
particles are occupied by smaller particles. These facies change discontinuously laterally in cross
sections and cause abrupt variations in hydraulic conductivity. Depositional and post-depositional
process of high-energy flow leads to complex sequences consisting of various facies, such that
geologic heterogeneity in alluvial coarse deposits is inherently large and trending.
165
166
A.4
167
Figure A. 6 Spatial trends in grain size and aquifer parameters in alluvial fan aquifers (Neton
et al. 1994)
168
A.5
Direct exchanges of water between streams (or rivers) and ground water occur in three basic ways.
Streams can gain water from ground-water inflow through their streambed, they can lose water through
their streambed to ground water, or they may do both: gaining water in some reaches and losing it in
others (Winter et al. 1998). Woessner (2000) classified four types of interactions between a stream and
groundwater: (1) gaining, where the groundwater flow into the groundwater; (2) losing, where the water
in the stream drains into the aquifer; (3) flow through, where the ground water flows into the stream on the
one side of the channel and out of the stream on the other side of the channel; and (4) parallel, where the
ground water flows in the aquifer beneath the stream and in the direction as the stream without entering or
leaving the stream.
Figure A. 8 Fluvial plain water and stream channel interactions showing channel cross
sections classified as: (a) gaining; (b) and (c) losing; (d) parallel-flow; (e) flow-through.
The stream is shaded. The water table and stream stage (thicker lines), groundwater flow
(arrows), and equipotential lines (dashed) are shown. Cross sections A, B, C, and E are
constructed parallel to ground water flow; D is perpendicular to ground water flow
(Woessner 2000)
169
A.6
The vertical component of hydraulic gradient at a given location can be measured by installing several
piezometers to different depths and measuring the hydraulic head at each piezometer. Ground water flows
in the direction of decreasing hydraulic head. The gradient direction indicates the potential for ground
water flow in that direction. However, the sufficient number of piezometrers is not installed in most
sites, so the information about vertical flows of groundwater actual flow direction is obtained in
scattered locations. One traditional approach is to plot well depth versus depth to static water level
from the groundwater table (Freeze and Cherry 1979). Under the assumption of a single topographic
region, the plot determines whether the region is a recharge or discharge area. However, recharge
and discharge areas in a basin are typically complex, and so it is not known a priori whether the site
can be treated as a single topographic region. In the plot method, the groundwater table is a common
datum surface, but the depth to the groundwater table is not uniform on such a regional scale, and
varies depending on topography, geology, and other factors.
Figure A. 9 Generalized plot of well depth versus depth to static water level
(Freeze and Cherry 1979)
170
A.7
Mann-Kendall test (Mann 1945; Kendall 1975) and Sens slope estimation (Sen 1968) are widely used
to test for randomness against trends in ecology, hydrology and climatology (e.g., Wen and Chen 2006;
Soderberg and Remy 2007). In Mann- Kendall tests, no trend or trend were e statistically tested whether
the null hypothesis H 0 of no trend, i.e. the observations are randomly ordered in time, were rejected or not.
Mann- Kendall test Static S value and its variance, VAR(S were respectively calculated by annual mean
water levels during the analysis periods. The standard test statistic Z value is obtained as:
(A.7.2)
A positive (negative) value of Z indicates an upward (downward) trend. To test for either an upward or
downward monotone trend (a two-tailed test) at level of significance, H0 was rejected if the |Z| > Z1 /2,
where Z1 /2 was obtained from the standard normal cumulative distribution tables. Sens slope was also
calculated using a linear trend model, and an estimate of the slope Qi is obtained by:
(A.7.3)
The statistical trend analysis can be archived using a free Excel macro (Salmi et al. 2002.)
171
A.8
172
complex nature.
(2) Statistical assumptions
Stationarity and ergodicity are also necessary to make stochastic approach useful (de Marsily
1986). Stationarity assumes that any statistical property of the geologic material (e.g., mean,
variance) is stationary in space, i.e. does not vary with a translation. In more rigorous terms, strong
staionarity means that all the probability distribution functions (pdf) are invariant under the
translation. In the case of geostatistics, week stationary is at least required where only the first two
moments are staionary: expected value is constant, i.e., E [Z(x] = m; covariance is not a function x
but a function only of the lag h, i.e. Cov(h) = E [(Z(x m) (Z(x+ h m) ].
Erdgodicity also implies that the unique realization available behaves in space with the same pdf
as the ensemble of possible realizations. In other words, by observing the variation in space of
property, it is possible to the pdf in the random function for all realizations. In most applications, it is
assumed that stochastic realizations match model statistics exactly. However, the assumption is
rarely achieved, nor should it. For example, multiple SGS realizations reproduce variogram, which
have discrepancies with the inherent variogram model. This is called ergodicity fluctuation (Deutsch
and Journel 1998). The ergodicity fluctuation depends on several factors such as simulation
argorithm, the density of condinitoning data, and the size of the simulation grid. Goovearts (1997)
also suggests that data that are subject to measurement error. The fluctuation of the realization
variogram is generally important when the range of the variogram model is large with respect with
the size of the simulated area. Furthermore, the ergodic hypothesis, i.e. the ensemble average is
statistically equivalent to the spatial average of a variable in one realization, is assumed to hold
(Dagan, 1997). In addition, the validity of the ergodic hypothesis has been questioned for highly
heterogeneous media (Sposito 1998).
(3) Variogram analysis
Spatial variability or auto-correlation structure of regional variable, Z(x), where x denotes the
point in the geometric space, is often characterized by a function of only a distance vector h = x1
x2; a covariance C(h) or a variogram (h) as:
C(h) = E[Z(x +h) Z(x)] m2
(A.8.1)
(A.8.2)
173
where m denotes constant mean, i.e. E[z(x)] = m. In the second-order stationary hypothesis; (1)
E[z(x)]= m (constant mean) and (2) the function of autocovariance only depends on the distance and
not on the points of reference, a relation between the covariance and the variogram is written as: (h)
= C(0) C(h). A variogram is also estimated under a less stringent hypothesis, called the intrinsic
hypothesis. Traditionally, the variogram has been used for modeling spatial variability rather than
the covariance although kriging systems are more easily solved with the covariance matrices
(Deutsch and Journel 1998).
B
a
Figure A. 10 (a) the pair of covariance and variance with the same parameters, (b)
parameters in variogram; is C is the sill (total variance), B is the nugget (variance of the
spatially uncorrelated portion of the field) and a is the practical range that is the distance over
which the field is correlated
The variogram models are determined by fitting to experimental variogram, which characterized the
spatial variability of the data. Experimental variogram is defined as a half of the average squared
difference between two attributed values as:
1 N d
2
zxid zxi
2N d i 1
(A.8.3)
where N(h) is the number of pairs, z(xi) is the target variable at the position xi. The separation vector h is
specified with some direction and lag distance h. Considering scattered measurements at scatter
locations (a) as shown in Figure A. 11, variogam model is performed using as following steps: (b) all
possible pairs of points are considered: (c) experimental variogram is produced by averaging over
given classes of lag distances between the measurement locations: (d) theoretical variogram models
are estimated through parameter optimizations to match experimental variogram. Anisotropy may
174
show different sensitivity with directions (Figure A. 12). This means that a large amount of
measurements are required to reveal the spatial variability, and to confirm the validity of variogam
model. For example, Journel und Huijbregts (1978) suggest that the number of pairs in a distance
class should not be less than about 30 to 50. In aquifer characterization, the number of hydraulic
properties is often unobtainable. Theoretical models are needed to satisfy statistical condition in
mathematics. The spherical model, the exponential model, and the Gaussian model are, respectively,
represented as:
B, h 0
3 h 1 h 3
h B CB
0 h a
2 a 2 a
C , h a
,h0
0
hBCB
3h ,h0
1exp
, h0
0
(A.8.5)
h 2
hBCB
,h0
1exp 3
(A.8.4)
(A.8.6)
where B is the nugget effect, C is the partial sill, the range a is the practical range, h is the separated
distance.
175
176
(4) Kriging
Kriging is an optimal interpolation procedure developed by Matheron as the theory of
regionalized variables. It was called kriging in recognition of Krige, which first proposed the origin
of this method. Kriging provides best unbiased linear estimates (BLUE) at any no measurement data
points, and also satisfies conditional condition, in which Kriging estimates at measurement points
are exactly equal to the measurements. Kriging also infers uncertainty (i.e. variance) besides
estimates at each desired points, while other deterministic estimation (e.g. IDW) cannot. In addition,
linear estimation procedure is easy to theoretically understood, and input in various available
software and GIS.
In geostatistics, the geological phenomenon is considered as a function of space and/or time. Let z(x)
represent any random function of the spatial coordinates z with measured values at N locations in space
z(xi) (i1,2,3, ... , N), and suppose that the value of the function z has to be estimated at the point x0. Each
measured value z1, z2, z 3, ... , z N contributes in part to the estimation of the unknown value z0 at location x0.
Taking this into account, and assuming a linear relation between z0 and zi, the estimated value of z0 can be
defined as (Ahmed et al. 2010):
N
(A.8.7)
where z * (x0) is the estimation of function z(x) at the point x0 and i are weighting coefficients. To
ensure an unbiased and optimal predictor the following two conditions need to be satisfied:
1. The expectant difference between the estimated value and the true (unknown value (the expected value
of the estimation error should be zero, i.e.
Ez* x0 zx0 0
(A.8.8)
(A.8.9)
i1
i 1
(A.8.10)
177
K2 x0 i i Ezxi zx j E z* x0 2 2 i Ezxi z* x0
N
i 1
j 1
i 1
(A.8.11)
i 1
j 1
i 1
(A.8.11)
i 1
j 1
i 1
i 1
Q i jC xi , x j C 02 iC xi , x0 22 i
i 1
(A.8.12)
Making partial differential equations of Q with respect to i and and equating them to zero
minimizes Equation (8.12) leading to the kriging equations:
C xi , x0 jC xi , x j ,i1,2,3..., N
j 1
(A.8.13)
j1
j 1
(A.8.14)
Substituting Equation (8.13) into Equation (8.14) yields the variance of the estimation error:
N
K2 x0 C0 iC xi , x0
i 1
(A.8.15)
The square root of this equation gives the standard deviation K (x0), which means that, with 95%
confidence, the true value will be within z*(x0) 2 K (x0).
If the covariance cannot be defined we can apply the kriging equation:
xi , x0 j xi , x j ,i1,2,3..., N
N
j 1
(A.8.16)
where (xi, xj) is the variogram between points xi and xj. The variance of the estimation error then
becomes
N
K2 x0 i xi , x0
i 1
(A.8.17)
178
Equation (8.16) in conjunction with (8.14) represents a set of (N + 1) linear equations with (N + 1)
unknowns, which on solution yield the coefficients i Once these are known, equations (8.7), (815)
and/or (8.17) can be evaluated. Compared other interpolation techniques (e.g., IDW), kriging is
advantageous because it considers (Journel and Hujibregts 1978); 1) the number and spatial
configuration of observation points; 2) the position of data points within the region of interest; 3) the
distances between the data points with respect to the area of interest; and 4) the spatial continuity of
the interpolated variable.
(5) Nonstationary problems
In nonstationary problems the mathematical expectation of z is no longer a constant: E[z(x)] =
m(x), and the variogram cannot be calculated directly from the data since m(x) is unknown.
Nonstationary problems in geostatistics have been discussed in the literature (e.g., de Marsily 1986;
Goovaerts 1997; Deutsch and Journel 1998; Wackernagel 2003).
There are several procedures for solving nonstationary problems. One simple assumption is
locally stationary as E[z(x)] = m(s) s A, where A is a certainly local field, and the variogram
stays isotropic. Kriging with moving neighborhood is conducted by setting a search radius less than
the distance in which the stationary assumption is satisfied.
Another way is to assume that the mathematical expectation m(x), i.e., global trend, is known or
estimated and to krige the residuals z(x) m(x) under the assumption that the residuals are stationary.
The mathematical expression might also be known for physical reasons, and then the constants of
this expression may be fitted on the model. A variety of kriging techniques are developed to
represent global drift using auxiliary variable, for example, porosity for hydraulic conductivity or
179
ground surface elevation for groundwater table elevation. The techniques are classified into
co-kriging based and regression-kriging based (Hengl 2007). Cokriging is reserved for linear
regression that also uses data defined on target variable and auxiliary variables. Cokriging needs
auxiliary variables at not all grid nodes. This is a main advantage to use cokriging, however
determination of cross-variogram is generally an effortful process. If auxiliary variables are obtained
at all grid nodes, and correlated with the target variable, the latter is generally preferred over the former. In
regression-kriging based approach, the regionalized variable is conceptualized as a sum of a
deterministic component (drift), a spatial auto-correlation component and a purely random
component as (Figure A. 14):
z (x) = m(s) + e (s) + e (s)
(A.8.18)
where z (x) is the target variable at the location s, e (s) is the spatial auto-correlation component, and
e (s) is the purely random component. Regression-kriging based approach is also divided into
Universal kriging (UK), kriging with external drift (KED), and Regression kriging (RK), of which
prediction processes are different but should give the same predictions and prediction error (Hengl 2007).
Universal kriging and kriging are performed in which kriging weights and drift parameters in both
kriging methods are simultaneously solved in the kriging system. The kriging system in both methods
requires in advance the covariance of residuals, which are obtained by subtracting the drift by the
variable, but this poses a conflict because drift parameters are determined after kriging.
Alternatively, the drift and residuals are fitted separately and then summed; this approach is
applied in various techniques such as residual kriging, kriging combined with regression, and
regression kriging. The applicability also has been indicated through comparison studies (e.g.,
Knotters et al. 1995; Hengl 2007; Nikroo et al. 2010). The formulas and parameters of drift are
firstly determined by ordinary least squares (OLS). Statistically, unbiased parameters of drift are also
determined by the iterative process of generalized least squares (GLS; Cressie 1993; Hengl 2007).
GLS is a sophisticated but laborious process. The first OLS drift should be satisfactory but slightly
inferior (Kitanidis 1993; Minasny and McBratney 2007). RK is direct and powerful method to
address non-stationary problem, but discrepancies between final estimates and measurements are
significantly caused if drift model is not enough satisfied. This means adequate model of drift is key
point to obtain accuracy of final estimates.
180
181
The process is repeatedly performed until simulation values at all desired nodes are determined. As a
result, it generates conditional realizations, and can generate very large random fields at irregular
grids as well as regular grids. Subsurface hydraulic conductivity frequently follows a log-normal
distribution, and in other words logK is normal (e.g. Freeze and Cherry 1979; Domenico and
Schwartz 1998). Sequential Gaussian simulation (SGS) is one method that seeks to overcome these
problems. The underlying principle is that the error variance at each estimated point, SK , often referred
to as the kriging variance, can be calculated from
k
SK x0 Cov(0) iCobx0xi
i 1
(A.8.19)
With SGS, the value estimated by Equation (8.7) is treated as the expectant. SGS then makes a
stochastic estimate of the value by adding a residual sampled from a Gaussian distribution of variance,
2 , as obtained from Equation (8.19) (hence Gaussian simulation). This value is then added to the
SK
collection of observations (i.e. the right-hand side of Equation (8.7)). This procedure is repeated
sequentially for each desired value of x0. It should be appreciated that the matrices in the kriging equation
also become increasingly larger with each estimate. Consequently, these calculations become very
computationally intensive and one needs special algorithms to make them feasible (Gomez-Hernandez
and Journel 1993).
182
A.9
If the properties of fluid and geologic material are independent of temperature, the
three-dimensional heat transport equation under the can be written as (Domenico and Schwartz
1998):
ke
2T
wcw
T
(Tq )
c
t
(A.8.20)
where T is the temperature; t is time; w and cw are, respectively, density and specific heat of the
fluid and and c are, respectively, density and specific heat of the geologic material (bulk values); q
is the seepage velocity or specific discharge vector; and ke is a term that includes the effective
thermal conductivity of the geologic material (bulk value).
Recent studies, however, suggest the importance of temperature dependence with respect to
solution robustness in groundwater modeling (Ma and Zheng 2010; Engeler et al. 2011). SEAWAT
used in this study also calculates groundwater flow and temperature transport with greater accuracy
by using iterative processes. The variable-density solute-transport and ground-water flow equation
(tensors and vectors shown in bold) is represented as:
h
0
C
0 K 0 h 0
z S s, 0 0
s q s
t
C t
0
(A.8.21)
where 0 is the fluid density [ML-3] at the reference concentration and reference temperature; is dynamic
viscosity; K0 is the hydraulic conductivity tensor of material saturated with the reference fluid; h0 is the
hydraulic head measured in terms of the reference fluid of a specified concentration and temperature (as the
reference fluid is commonly freshwater); Ss,0 is the specific storage, defined as the volume of water released
from storage per unit volume per unit decline of h0; t is time; is porosity; C is salt concentration [ML-3];
and q's is a source or sink [T-1] of fluid with density s.
Equation (8.21) represents solute transport, not only heat transport under the variable condition of
density and viscosity. The fluid density and viscosity can be calculated using concentrations from one or
more solute species. These equations of state relate the density and viscosity terms in equation 1 to one or
more of the MT3DMS species concentrations (Ck). Equation (8.21) can also be used to represent heat
transport. In this case, one of the MT3DMS solute species is used to represent temperature, and the
mathematical terms describing heat transport are converted into forms analogous to those used in Equation
183
(8.21). By allowing one or more of the MT3DMS species to affect fluid density and viscosity,
variable-density ground-water flow can be coupled with simultaneous solute and heat transport.
SEAWAT utilized the density function of temperature as:
NS
k 1 C k
where
can be calculated from the volumetric expansion coefficient for pressure, P, using:
02 g P
(A.8.23)
Dynamic viscosity is considered here to be a function of only temperature and solute concentration, which
is the typical approach, and neglects the weak dependence of viscosity on fluid pressure. In SEAWAT, an
alternative equation for dynamic viscosity has been implemented:
248.37
(A.8.24)
184
Appendix B
Comparison between calculated and measured values of groundwater temperature in study cases as
Case1a: Trending heterogeneity and isotropy
Case1b: Trending heterogeneity and anisotropy
Case1c: Trending heterogeneity and large anisotropyisotropy
Case2a: Stationary heterogeneity and isotropy
Case2b: Stationary heterogeneity and anisotropy
Case2c: Stationary heterogeneity and large anisotropyisotropy
Case3: Case1b with tenth value in vertical hydraulic conductivity of riverbed
Case4: Case1b with twice value in river width
Case5: Case1b with isotropy condition in thermal diffusivity
Case6: Case1b with tenth value in ranges in variograms
Case7: Case1b with no-temperature dependence of fluid
Case8a: Block averaged of Case1a
Case8b: Block averaged of Case1b
185
Case 1a
BW3-1: Z= 15m
BW3-1: Z= 30m
BW3-1: Z= 60m
no. 16
15
15
15
10
0
0
Temperature
20
10
50
100
150
day
200
0
0
250
10
50
100
150
200
0
0
250
BW3-4:Z=10.5m
15
Temperature
15
Temperature
15
10
100
150
day
200
250
0
0
250
no. 16
20
200
no. 16
20
10
150
BW3-5: Z= 4m
no. 16
50
100
day
20
0
0
50
day
BW3-2: Z= 5m
Temperature
no. 16
20
Temperature
Temperature
no. 16
20
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
186
Case 1b
BW3-1: Z= 15m
BW3-1: Z= 30m
no. 4
no. 47
20
15
Temperature
10
10
50
100
150
day
200
0
0
250
50
150
200
0
0
250
no. 4
no. 47
no. 4
no. 47
200
250
250
no. 4
no. 47
20
Temperature
10
0
0
200
15
150
150
BW3-5: Z= 4m
20
Temperature
10
day
100
day
15
100
50
BW3-4: Z= 10.5m
15
Temperature
100
day
20
50
10
BW3-2: Z= 5m
0
0
no. 4
no. 47
20
15
Temperature
Temperature
no. 4
no. 47
20
15
0
0
BW3-1: Z= 60m
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
187
Case 1c
BW3-1: Z= 15m
BW3-1: Z= 30m
BW3-1: Z= 60m
15
15
15
10
0
0
Temperature
20
10
50
100
150
day
200
0
0
250
10
50
100
150
200
0
0
250
BW3-4: Z= 10.5m
15
Temperature
15
Temperature
15
10
100
150
day
200
250
0
0
250
no. 13
20
200
no. 13
20
10
150
BW3-5: Z= 4m
no. 13
50
100
day
20
0
0
50
day
BW3-2: Z= 5m
Temperature
no. 13
no. 13
20
Temperature
Temperature
no. 13
20
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
188
Case 2a
BW3-1: Z= 30m
20
15
15
15
10
0
0
Temperature
20
10
50
100
150
day
200
0
0
250
10
50
100
150
200
0
0
250
BW3-4: Z= 10.5m
15
15
Temperature
15
Temperature
20
10
10
100
150
day
200
250
0
0
150
200
250
200
250
BW3-5: Z= 4m
20
50
100
day
20
0
0
50
day
BW3-2: Z= 5m
Temperature
BW3-1: Z= 60m
20
Temperature
Temperature
BW3-1: Z= 15m
10
50
100
150
day
200
250
0
0
50
100
day
150
189
Case 2b
BW3-1: Z= 30m
20
15
15
15
10
0
0
Temperature
20
10
50
100
150
day
200
0
0
250
10
50
100
150
200
0
0
250
BW3-4: Z= 10.5m
15
15
Temperature
15
Temperature
20
10
10
100
150
day
200
250
0
0
150
200
250
200
250
BW3-5: Z= 4m
20
50
100
day
20
0
0
50
day
BW3-2: Z= 5m
Temperature
BW3-1: Z= 60m
20
Temperature
Temperature
BW3-1: Z= 15m
10
50
100
150
day
200
250
0
0
50
100
day
150
190
Case 2c
BW3-1: Z= 30m
20
15
15
15
10
0
0
Temperature
20
10
50
100
150
day
200
0
0
250
10
50
100
150
200
0
0
250
BW3-4: Z= 10.5m
15
15
Temperature
15
Temperature
20
10
10
100
150
day
200
250
0
0
150
200
250
200
250
BW3-5: Z= 4m
20
50
100
day
20
0
0
50
day
BW3-2: Z= 5m
Temperature
BW3-1: Z= 60m
20
Temperature
Temperature
BW3-1: Z= 15m
10
50
100
150
day
200
250
0
0
50
100
day
150
191
Case 3
BW3-1: Z= 15m
BW3-1: Z= 30m
BW3-1: Z= 60m
15
15
15
10
0
0
Temperature
20
Temperature
Temperature
20
10
50
100
150
day
200
0
0
250
10
50
100
150
200
0
0
250
BW3-4: Z= 10.5m
15
Temperature
15
Temperature
15
10
100
150
day
200
250
0
0
250
no. 4
20
200
no. 4
20
10
150
BW3-5: Z= 4m
no. 4
50
100
day
20
0
0
50
day
BW3-2: Z= 5m
Temperature
no. 4
no. 4
no. 4
20
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
192
Case 4
BW3-1: Z= 30m
BW3-1: Z= 15m
BW3-1: Z= 60m
no. 63
15
15
15
10
0
0
Temperature
20
10
50
100
150
day
200
0
0
250
10
50
100
150
200
0
0
250
BW3-4: Z= 10.5m
15
Temperature
15
Temperature
15
10
100
150
day
200
250
0
0
250
no. 63
20
200
no. 63
20
10
150
BW3-5: Z= 4m
no. 63
50
100
day
20
0
0
50
day
BW3-2: Z= 5m
Temperature
no. 63
20
Temperature
Temperature
no. 63
20
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
193
Case 5
BW3-1: Z= 15m
BW3-1: Z= 30m
no. 4
no. 7
20
15
Temperature
10
10
50
100
150
day
200
0
0
250
50
150
200
0
0
250
no. 4
no. 7
no. 4
no. 7
200
250
250
no. 4
no. 7
20
Temperature
10
0
0
200
15
150
150
BW3-5: Z= 4m
20
Temperature
10
day
100
day
15
100
50
BW3-4: Z= 10.5m
15
Temperature
100
day
20
50
10
BW3-2: Z= 5m
0
0
no. 4
no. 7
20
15
Temperature
Temperature
no. 4
no. 7
20
15
0
0
BW3-1: Z= 60m
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
194
Case 6
BW3-1: Z= 30m
20
15
15
15
10
0
0
Temperature
20
10
50
100
150
day
200
0
0
250
10
50
100
150
200
0
0
250
BW3-4: Z= 10.5m
15
15
Temperature
15
Temperature
20
10
10
100
150
day
200
250
0
0
150
200
250
200
250
BW3-5: Z= 4m
20
50
100
day
20
0
0
50
day
BW3-2: Z= 5m
Temperature
BW3-1: Z= 60m
20
Temperature
Temperature
BW3-1: Z= 15m
10
50
100
150
day
200
250
0
0
50
100
day
150
195
Case 7
BW3-1: Z= 30m
BW3-1: Z= 15m
no. 7
no. 63
20
15
Temperature
10
10
50
100
150
day
200
0
0
250
50
150
200
0
0
250
no. 7
no. 63
no. 7
no. 63
200
250
250
no. 7
no. 63
20
Temperature
10
0
0
200
15
150
150
BW3-5: Z= 4m
20
Temperature
10
day
100
day
15
100
50
BW3-4: Z= 10.5m
15
Temperature
100
day
20
50
10
BW3-2: Z= 5m
0
0
no. 7
no. 63
20
15
Temperature
Temperature
no. 7
no. 63
20
15
0
0
BW3-1: Z= 60m
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
196
Case 8a
BW3-1: Z= 30m
BW3-1: Z= 15m
no. 1
no. 8
no. 70
no. 78
no. 92
20
15
Temperature
10
10
50
100
150
day
200
0
0
250
50
150
200
0
0
250
no. 1
no. 8
no. 70
no. 78
no. 92
no. 1
no. 8
no. 70
no. 78
no. 92
200
250
250
no. 1
no. 8
no. 70
no. 78
no. 92
20
Temperature
10
0
0
200
15
150
150
BW3-5: Z= 4m
20
Temperature
10
day
100
day
15
100
50
BW3-4: Z= 10.5m
15
Temperature
100
day
20
50
10
BW3-2: Z= 5m
0
0
no. 1
no. 8
no. 70
no. 78
no. 92
20
15
Temperature
Temperature
no. 1
no. 8
no. 70
no. 78
no. 92
20
15
0
0
BW3-1: Z= 60m
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
197
Case 8b
BW3-1: Z= 30m
10
15
0
0
10
100
150
day
200
0
0
250
50
150
200
10
150
100
200
250
no. 1
no. 7
no. 8
no. 9
no. 17
no. 18
no. 25
no. 30
no. 33
no. 42
no. 44
no. 63
no. 65
no. 78
15
10
0
0
150
200
250
BW3-5: Z= 4m
day
50
day
20
Temperature
no. 1
no. 7
no. 8
no. 9
no. 17
no. 18
no. 25
no. 30
no. 33
no. 42
no. 44
no. 63
no. 65
no. 78
100
0
0
250
BW3-4: Z= 10.5m
15
Temperature
100
day
20
50
10
BW3-2: Z= 5m
0
0
15
50
no. 1
no. 7
no. 8
no. 9
no. 17
no. 18
no. 25
no. 30
no. 33
no. 42
no. 44
no. 63
no. 65
no. 78
20
no. 1
no. 7
no. 8
no. 9
no. 17
no. 18
no. 25
no. 30
no. 33
no. 42
no. 44
no. 63
no. 65
no. 78
20
15
Temperature
Temperature
15
no. 1
no. 7
no. 8
no. 9
no. 17
no. 18
no. 25
no. 30
no. 33
no. 42
no. 44
no. 63
no. 65
no. 78
20
Temperature
no. 1
no. 7
no. 8
no. 9
no. 17
no. 18
no. 25
no. 30
no. 33
no. 42
no. 44
no. 63
no. 65
no. 78
20
BW3-1: Z= 60m
Temperature
BW3-1: Z= 15m
10
50
100
150
day
200
250
0
0
50
100
day
150
200
250
198
Appendix C
Simulation results of selected optimum in each study case at the last period (ti=
264day), Blue lines, red lines, and green arrows show groundwater head contours,
temperature contours, and flow velocity vectors (a: directions and b: magnitudes),
respectively.
199
Case 1a
Realization no.16; T=50days; Maximum velocity = 260 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
200
Distance (m)
Case 1a
Realization no.16; T=150days; Maximum velocity = 240 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
201
Distance (m)
Case 1a
Realization no.16; T=264days; Maximum velocity = 270 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
202
Distance (m)
Case 1b
Realization no.4; T=50days; Maximum velocity = 130 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
203
Distance (m)
Case 1b
Realization no.4; T=150days; Maximum velocity = 110 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
204
Distance (m)
Case 1b
Realization no.4; T=264days; Maximum velocity = 140 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
205
Distance (m)
Case 1c
Realization no.13; T=50days; Maximum velocity = 68 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
206
Distance (m)
Case 1c
Realization no.13; T=150days; Maximum velocity = 55 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
Elevation (m asl)
BW03-4
Distance (m)
Elevation (m asl)
River
207
Distance (m)
BW03-5
Case 1c
Realization no.13; T=264days; Maximum velocity = 63 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
208
Distance (m)
Case 3
Realization no.4; T=50days; Maximum velocity = 100 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
209
Distance (m)
Case 3
Realization no.4; T=150days; Maximum velocity = 120 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
210
Distance (m)
Case 3
Realization no.4; T=264days; Maximum velocity = 160 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
Elevation (m asl)
BW03-4
Distance (m)
River
Elevation (m asl)
BW03-5
211
Distance (m)
Case 4
Realization no.63; T=50days; Maximum velocity = 310 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
212
Distance (m)
Case 4
Realization no.63; T=150days; Maximum velocity = 270 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
213
Distance (m)
Case 4
Realization no.63; T=264days; Maximum velocity = 290 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
214
Distance (m)
Case 5
Realization no.4; T=50days; Maximum velocity = 130 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
215
Distance (m)
Case 5
Realization no.4; T=150days; Maximum velocity = 110 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
216
Distance (m)
Case 5
Realization no.4; T=264days; Maximum velocity = 140 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
217
Distance (m)
Case 7
Realization no.63; T=50days; Maximum velocity = 320 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
218
Distance (m)
Case 7
Realization no.63; T=150days; Maximum velocity = 280 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
219
Distance (m)
Case 7
Realization no.63; T=264days; Maximum velocity = 300 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
220
Distance (m)
Case 8a
Realization no.1; T=50days; Maximum velocity = 97 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
221
Distance (m)
Case 8a
Realization no.1; T=150days; Maximum velocity = 87 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
222
Distance (m)
Case 8a
Realization no.1; T=264days; Maximum velocity = 100 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
223
Distance (m)
Case 8b
Realization no.17; T=50days; Maximum velocity = 640 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
224
Distance (m)
Case 8b
Realization no.17; T=150days; Maximum velocity = 390 m/d
BW03-2
BW03-1
BW03-2
BW03-1
River
BW03-5
Elevation (m asl)
BW03-4
Distance (m)
River
Elevation (m asl)
BW03-4
225
Distance (m)
BW03-5
Case 8b
Realization no.17; T=264days; Maximum velocity = 650 m/d
BW03-2
BW03-1
BW03-4
BW03-2
BW03-1
River
BW03-5
River
BW03-5
Elevation (m asl)
BW03-4
Elevation (m asl)
Distance (m)
226
Distance (m)
References
Aboufirassi M and Mario MA (1983) Kriging of water levels in the Souss aquifer, Morocco.
Mathematical Geology, doi: 10.1007/BF01031176
Abriola LM and Pinder GF (1982) Calculation of velocity in three space dimensions from hydraulic
head measurements. Ground Water, 20(2), 205-213, doi: 10.1111/j.1745-6584.1982.tb02752.x
Ahmed S, Sarah S, Nabi A and Owais S (2010) Performing unbiased groundwater modeling:
application of the theory of regionalized variables, In Groundwater Modeling in Arid and
Semi-Arid Areas (Ed. Wheater HS, Mathias, SA and Li X). Cambridge University Press
Anderson MP (1989) Hydrogeologic facies models to delineate large-scale spatial trends in glacial
and glaciofluvial sediments. GSA Bulletin 101:501-511, doi:10.1130/0016-7606
Anderson MP and Woessner WW (1992) Applied Groundwater Modeling Simulation of Flow and
Advective Transport. Academic press
Anderson MP (2005) Heat as a ground water tracer, Ground Water, 43, 951-968
Anderson MP (2010) Heat transport in groundwater systems in Japan. Journal of Ground Water
Hydrology, 52(4), 355369
ASCE (American Society of Civil Engineers) (2008) Standard guideline for fitting saturated
hydraulic conductivity using probability density functions. ASCE
Bear J (1972) Dynamics of Fluids in Porous Media. Dover
Bear J (1979) Hydraulics of Groundwater. McGraw-Hill
Bennett GL, Weissmann GS, Baker GS and Hyndmann DW (2006) Regional-scale assessment of a
sequence-bounding paleosol on fluvial fans using ground-penetrating radar, eastern San
Joaquin Valley, California. Bulletin of the Geological Society of America, 118, 724732
Bower H (1978) Groundwater Hydrology, McGraw-Hill
Bull WB (1968) Alluvial fans. Journal of Geologic Education, 16, 101-106.
Burrough PA and McDonnell RA (1998) Principles of Geographical Information Systems. Oxford
University Press
Cardenas MB and Jiang XW (2010) Groundwater flow, transport, and residence times through
topography-driven basins with exponentially decreasing permeability and porosity. Water
Resoures Research 46, W11538, doi :10.1029/2010WR009370
227
Carrera J and Mathias SA (2010) Groundwater flow and transport, In Groundwater Modeling in
Arid and Semi-Arid Areas (Ed. Wheater HS, Mathias, SA and Li X). Cambridge University
Press
Cehrs D (1979) Depositional control of aquifer characteristics in alluvial fans, Fresno Country,
Calfornia. Geologic Society of America Bulletin, 90(2), 1282-1309
Chen X (2011) Depth-dependent hydraulic conductivity distribution patterns of a streambed.
Hydrological Processes, 25, 278-287, doi:10.1002/hyp.7844
Cheong JY, Hamm SY, Kim HS, Ko EJ, Yang K and Lee JH (2008) Estimating hydraulic
conductivity using grain-size analyses, aquifer tests, and numerical modeling in a riverside
alluvial
system
in
South
Korea.
Hydrogeology
Journal,
16,
1129-1143,
doi:
10.1007/s10040-008-0303-4
Chung JW and JD Rogers (2012) Interpolations of groundwater table elevation in dissected uplands,
Ground Water, 50(4), 598-607, doi: 10.1111/j.1745-6584.2011.00889.x
Chils JP and Delfiner P (1999) Geostatistics: Modeling Spatial Uncertainty. John Wiley and Sons
Constantz J (2008) Heat as a tracer to determine streambed water exchanges, Water Resources
Research, 44, W00D10, doi: 10.1029/2008WR006996
Constantz J, Niswonger NRG and Stewart AE (2008) Analysis of Temperature Gradients to
Determine Stream Exchanges with Ground Water, In Field Techniques for Estimating Water
Fluxes Between Surface Water and Ground Water (Ed. Rosenberry DO and LaBaugh JW).
USGS
Cressie NAC (1993) Statistics for Spatial Data Revised Edition. John Wiley and Sons
Dagan G (1997) Stochastic modeling of flow and transport: the broad perspective, In Subsurface
Flow and Transport: A Stochastic Approach (Ed. Dagan G and Neuman SP). Cambridge
University Press
Daimaru H (1989) Holocene evolution of the Toyohira River alluvial fan and distal floodplain,
Hokkaido, Japan. Geophysical Review of Japan, 62(A-8), 589-603, In Japanese
Daimaru H (2003) Toyohira River alluvial fan, In Regional Geomorphology of the Japanese Islands
vol.2 Geomorphology of Hokkaido (Ed. Koaze T, Nogami M, Ono Y, Hirakawa K).
University of Tokyo Press, In Japanese
228
229
doi: 10.1007/s10040-007-0187-8
Ferreira JT, Ritzi Jr. RW and Dominic DF (2010) Measuring the permeability of open-framework
gravel. Ground Water, 48, 593-597, doi: 10.1111/j.1745-6584.2010.00675.x
Fetter CW (2001) Applied hydrogeology 4th edn. Prentice-Hall
Fleckenstein JH, Niswonger RG and Fogg GE (2006) River-aquifer interactions, geologic
heterogeneity, and low-flow management. Ground Water, 44(6), 837-852, doi:
10.1111/j.1745-6584.2006.00190.x
Freeze RA and Cherry JA (1979) Groundwater. Prentice-Hall
Frei S, Fleckenstein JH, Kollet SJ and Maxwell RM (2009) Patterns and dynamics of riveraquifer
exchange with variably-saturated flow using a fully-coupled model. Journal of Hydrology,
375, 383393, doi: 10.1016/j.jhydrol.2009.06.038
Fritz BG and Mackley RD (2010) A wet/wet differential pressure sensor for measuring vertical
hydraulic gradient. Ground Water, 48(1), 117-121, doi: 10.1111/j.1745-6584.2011.00789.x
(2010) 1 (),
443-447
(1928) ()35(418) 382-387
(1997) ,
Gee GW, Hillel D (1988) Groundwater recharge in arid regions: Review and critique of estimation
methods. Hydrological Processes, 2(3), 255266, doi: 10.1002/hyp.3360020306
Genereux DP, Leahy S, Mitasova H, Kennedy CD and Corbett DR (2008) Spatial and temporal
variability of streambed hydraulic conductivity in West Bear Creek, North Carolina, USA.
Journal of Hydrology, 358, doi: 10.1016/j.jhydrol.2008.06.017
Goldschneider AA, Haralampides KA and MacQuarrie KTB (2007) River sediment and flow
characteristics near a bank filtration water supply: implications for riverbed clogging. Journal
of Hydrology, 344, 55-69, doi: 10.1061/j.jhydrol.2007.06.031
Gomez-Hernandez JJ and Journel A (1993) Joint sequential simulation of multigaussian fields, In
Geostatistics Tria 92. Springer
Goovaerts P (1997) Geostatistics for Natural Resources Evaluation. Oxford University Press
230
Gray WG and Pinder GF (1976) An Analysis of the Numerical Solution of the Transport Equation.
Water Resources Research, 12, 547-555
Guo W and Bennett GD (1998) SEAWAT version 1.1: A computer program for simulations of
groundwater flow of variable density. Missimer International, Inc.
Guo W and Langevin CD (2002) Users Guide to SEAWAT: A computer program for simulation of
three-dimensional variable-density ground-water flow. USGS
Hails JR (1976) Compaction and diagenesis of very coarse-grained sediments, In Compaction of
Coarse-Grained Sediments, I (Ed. Chilingarian GV and Wolf KH). Elsevier
Harbaugh AW, Banta ER, Hill MC and McDonald MG (2000) MODFLOW-2000, The US
Geological Survey modular ground-water model-User Guide to modularization concepts and
the ground-water flow process. USGS
Harte PT and Kiah RG (2009) Measured river leakages using conventional streamflow techniques:
the case of Souhegan River, New Hampshire, USA. Hydrogeology Journal, 17, 409-424, doi:
10.1007/s10040-008-0359-1
Hatanaka M, Uchida A, Taya Y, Takehara N, Hagisawa T, Sakou N and Ogawa S (2001)
Permeability Characteristics of High-QualityUndisturbed Gravelly Soils Measured in
Laboratory Tests, Soils and Foundations, 41(3), 45-55
Hazen A (1911) Discussion: Dams on sand foundations. Transactions, ASCE 73:199-203
Healy RW (2010) Estimating Groundwater Recharge. Cambridge University Press
Heinz J, Kleineidam S, Teutsch G and Aigner T (2003) Heterogeneity patterns of Quaternary
glaciofluvial gravel bodies (SW-Germany): application to hydrogeology. Sedimentary
Geology, 158, 1-23
Hess KM, Wolf SH and Celia MA (1992) Large-scale natural gradient tracer test in sand and gravel,
Cape Cod, Massachusetts: 3. Hydraulic conductivity variability and calculated
macrodispersivities. Water Resources Research, 28, 2011-2027, doi: 10.1029/92WR00668
Hengel T (2007) A Practical Guide to Geostatistical Mapping of Environmental Variables. JRC
Science and Technical Reports, EUR 22904 EN-2007, Office for Official Publications of the
European Communities
Hoeksema RJ, Clapp RB, Thomas AL, Hunley AE, Farrow ND and Dearstone KC (1989)
231
Cokriging model for estimation of water table elevation. Water Resources Research, 25(3),
429-438
(2006a)
(2006b)
(2008)
(1992) (15) 3-2.
34(1)31-40
Hu SG, Yamazaki M, Sataka H, Akiyama T and Hirano R (2001) Tephrochronological
reconsideration on Shikotsu Pumice Flow and Fall Deposits, Hokkaido, Japan. Earth Science,
55(3), 145-156, In Japanese
Hu SG, Kobayashi T, Okuda E, Oishi A, Saito M, Kayaki, T and Miyazaki S (2010a) Alluvial
fans-importance and relevance: a review of studies of Research group on Hydro-environments
around alluvial fans in Japan, In Groundwater Response to Changing Climate (ed. Taniguchi
M and Holman IP). 131-139, CRC Press, doi: 10.1201/b10530-13
Hu SG, S Miyajima, D Nagaoka, K Koizumi and K Mukai (2010b) Study on the relation between
groundwater and surface water in Toyohira-gawa alluvial fan, Hokkaido, Japan, In
Groundwater Response to Changing Climate (ed. Taniguchi M and Holman IP). 141-158,
CRC Press, doi: 10.1201/b10530-13
Hubbert MK (1940) The theory of groundwater motion. Journal of Geology, 48, 785-944
Huyakorn PS and Pinder GF (1983) Computational Methods in Subsurface Flow. Academic Press
Hvorslev MJ (1951) Time lag and soil permeability in ground water observations. U.S. Army
Waterways Experiment Station, Bulletin, 36
Institute of Civil Engneers (ICE; 1976) Manual of applied geology for engineers. ICE
Ingebritsen SE, Sanford WE and Neuzil CE (2006) Groundwater in geologic processes 2nd edn.
Cambridge university press
Ishida M, Soya T and Suda Y (1980) Geological map of Japan 1:200,000 Sapporo. Geological
survey of Japan, NK-54-14
International Organization for Standardization (ISO; 2007) Hydrometry - Measurement of Liquid
Flow in Open Channels Using Current Meters or Floats Fourth edition. ISO
232
Jang CS and Liu CW (2005) Contamination potential of nitrogen compounds in the heterogeneous
aquifers of the Choushui River alluvial fan, Taiwan, Journal of Contaminant Hydrology, 79,
135-155
Jessop AM (1990) Thermal geophysics. Elsevier
JGS (Japan Geotechnical Society) (2004) Method for determination of hydraulic properties of
aquifer in single borehole, In: Japanese standards for geotechnical and geoenvironmental
investigation methods. Japan Geotechnical Society
Jiang XW, Wan L, Wang XS, Ge S and Liu J (2009) Effect of exponential decay in hydraulic
conductivity with depth on regional groundwater flow. Geophysical Research Letters, 36,
L24402, doi: 10.1029/2009GL041251
Jumikis AR (1977) Thermal geotechnics. Rutgers university press
Jiang XW, Wang XS and Wan L (2010) Semi-empirical equations for the systematic decrease in
permeability with depth in porous and fractured media. Hydrogeology Journal, 18, 839850,
doi: 10.1007/s10040-010-0575-3
Johnson LE (2009) Geographic Information Systems in Water Resources Engneering. CRC Press
Journel AG, Gundeso R, Gringarten E and Yao T (1998) Stochastic modelling of a fluvial reservoir:
a comparative review of algorithms. Journal of Petroleum Science and Engineering, 21,
95-121.
Journel AG and Huijbregts CJ (1978) Mining geostatistics. Academic press
Jussel P, Stauffer F and Dracos T (1994) Transport modeling in heterogeneous aquifers: 1. statistical
description and numerical generation of gravel deposits. Water Resources Research, 30,
1803-1817
Kamann PJ, Ritzi RW, Dominic DF and Conrad CM (2007) Porosity and permeability in sediment
mixtures. Ground Water, 45, 429-438, doi: 10.1111/j.1745-6584.2007.00313.x
Kasper JW, Denver JM, McKenna TE and Ullman WJ (2010) Simulated impacts of artificial
groundwater recharge and discharge on the source area and source volume of an Atlantic
Coastal Plain stream, Delaware, USA. Hydrogeology Journal, 18, 1855-1866, doi:
10.1007/s10040-010-0641-x
Katakai T, Ogita A, Sumi T and Tsujimoto T (2006) Experimental study on the clogging soil
233
porosity on boundary area between water edge and river bed, Proceedings of Hydraulic
Engineering, Japan Society of Civil Engneering, (50), 241-246, In Japanese
Kato M, Fukatsukawa K, Kikuchi J, Matsumoto K (1995) Subsurface geology and sedimentary
structure of alluvium in Sapporo city. Proceedings of the thirty-five Conference of Hokkaido
Branch, Japan Geotechnical Society Engineering, 82-89, Japanese Geotechniacal Society, In
Japanese
Kawata E and Obara T (1958) Underground Water in the Top of the Sapporo-Fan. Report of the
Geological Survey of Hokkaido, 19, 30-31, Geological Survey of Hokkaido, In Japanese
Kayane I and Yamamoto S (1971) Hydrologic Cycle in Alluvial Fan. Kokon-shoin, In Japanese
(1991)
Kendall MG (1975) Rank Correlation Methods. Charles Griffin
Kitanidis PK (1993) Generalized covariance functions in estimation. Mathematical Geology, 25(5),
525-540
Knotters M, Brus DJ and Oude Voshaar JH (1995) A comparison of kriging, co-kriging and kriging
combined with regression for spatial interpolation of horizon depth with censored
observations. Geoderma, 67, 227-246
Kodama Y and Inokuchi M (1986) Changes in the sedimentary fabric of surface deposits along the
lower course of the river Waterase. Bulletin of the Environmental Research Center, 10, 67-79
In Japanese
(2006) .
Koizumi K, Mukai K, Konishi H and Hu SG (2008) Change of Water Balance by Global Warming
in the Toyohira-gawa Alluvial Fan. In: Hydro-environments of Alluvial Fans in Japan (ed.
Secretariat of RHF), 13-28, Research group on Hydro-environment around alluvial Fans
Koltermann CE and Gorelick SM (1995) Fractional packing model for hydraulic conductivity
derived sediment mixtures. Water Resources Research, 31, 3283-3297
Koltermann CE and Gorelick SM (1996) Heterogeneity in sedimentary deposits: a review of
structure-imitating, process-imitating, and descriptive approaches. Water Resources Research
32(9), doi:10.1029/96WR00025
(2006)
234
, e- 2006
Kresic N (2007) Hydrogeology and groundwater modeling 2nd edn. CRC Press
Kundzewicz ZW, Mata LJ, Arnell NW, et al. (2007) Freshwater resources and their management. In
Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group
II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, ed.
Parry ML, Canziani OF, Palutikof JP, van der Linden PJ, aand Hanson CE, 173-210,
Cambridge University Press
Langevin CD, Shoemaker WB and Guo W (2003) MODFLOW-2000, the US Geological Survey
modular ground-water model-documentation of the SEAWAT-2000 version with the variable
density flow process (VDF) and the Integrated MT3DMS Transport Process (IMT). Open File
Report 03-426, USGS
Langevin CD, Thorne Jr DT, Dausman AM, Sukop MC and Guo W (2008) SEAWAT version 4: a
computer program for simulation of multi-species solute and heat transport, USDI/USGS
Lamotte JL and Delay F (1997) On the stability of the 2D interpolation algorithms with uncertain
data. Mathematics and Computers in Simulation, 43, 183-201
Lapham WW (1989) Use of Temperature Profiles Beneath Streams to Determine Rates of Vertical
Ground-Water Flow and Vertical Hydarulic Conductivit. Water-Supply Paper 2337, USGS
Lecce SA (1990) The alluvial fan problems, In Alluvial fans: a field approach (ed. Rachocki AH and
Church M). John Wiley and Sons
Lee JY, Yi MJ and Hwang D (2005) Dependency of hydrologic responses and recharge estimates on
water-level monitoring locations within a small catchment. Geosciences Journal, 9(3),
277-286, doi: 10.1007/BF02910588
Lee SY, Carle SF and Fogg GE (2007) Geologic heterogeneity and a comparison of two
geostatistical models: Sequential Gaussian and transition probability-based geostatistical
simulation. Advances in Water Resources, 30, doi: 10.1016/j.advwatres.2007.03.005
Lunt IA, Bridge JS and Tye RA (2004) A quantitative, three-dimensional depositional model of
gravelly braided rivers. Sedimentology, 41, 377-414, doi: 10.1111/j.1365-3091.2004.00627.x
Lunt IA and Bridge JS (2007) Formation and preservation of open-framework gravel strata in
unidirectional flows. Sedimentology, 54, 71-87, doi: 10.1111/j.1365-3091.2006.00829.x
235
Luo W, Grudzinski B and Pederson D (2011) Estimating hydraulic conductivity for the Martian
subsurface based on drainage patterns-a case study in the Mare Tyrrhenum Quadrangle.
Geomorphology, 125, 414-420, doi: 10.1016/j.geomorph.2010.10.018
Ma R and Zheng C (2010) Effects of Density and Viscosity in Modeling Heat as a Groundwater
Tracer. Ground Water, 48(3), 380-389, doi: 10.1111/j.1745-6584.2009.00660.x
Major JJ (1997) Depositional processes in large-scale debris-flow experiments. The Journal of
Geology, 105, 345-366, doi: 10.1086/515930
Major JJ (2000) Gravity-driven consolidation of granular slurries: implications for debris-flow
deposition and deposit characteristics. Journal of Sedimentary Research, 70, 64-83, doi:
1073-130X/00/070-64/$03.00
Mann HB (1945) Non-parametric tests against trend. Econometrica, 13(2), 245-259
Manning CE and Ingebritsen SE (1999) Permeability of the continental crust: implications of
geothermal data and metamorphic systems. Reviews of Geophysics, 37, 127-150, doi:
10.1029/1998RG900002
Marion A, Packman AI, Zaramella M and Bottacin-Busolin A (2008) Hyporheic flows in stratified
beds. Water Resoures Research, 44, W09433, doi: 10.1029/2007WR006079
Matsumoto N and Yamaguchi Y (1991) Interaction between stress and permeability in sand and
gravel deposit. Journal of Geotechnical Engineering, 430, 59-67, In Japanese
McElwee CD, Butler Jr.JJ and Healey JM (1991) A New sampling system for obtaining relatively
undisturbed samples of unconsolidated coarse sand and gravel. Ground Water Monitoring &
Remediation, 11, 182-191, doi: 10.1111/j.1745-6592.1991.tb00390.x
McIlvride WA and Rector BM (1988) Comparison of short-and long-screen monitoring wells in
alluvial sediments. In Proceedings of the Second National Outdoor Action Conference on
Aquifer Restoration, Ground Water Monitoring and Geophysical Methods, 375-390, National
Water Well Association
Mercer JW and Faust CR (1981) Groundwater modeling. National Water Well Association
Miall AD (1992) Alluvial deposits. In Facies models response to sea level change (Ed. Walker RG
and James NP), Geological Association of Canada
Minasny B and McBratney AB (2007) Spatial prediction of soil properties using EBLUP with the
236
(1983)
47-63
National and Regional Policy Bureau (2007) Water Well Data in Japan. Geographical information
system portal site (http://www.gis.go.jp/contents/list/detail/000005.html), Accessed 25 June
2010.
Neton MJ, Dorsch J, Olson CFD and Young SC (1994) Architecture and directional scales of
heterogeneity in alluvail-fan aquifers. Journal of Sedimentary Research, 2, 245-257
Nikroo L, Kompani-Zare M, Sepaskhah AR and Shamsi SRF (2010) Groundwater depth and
elevation interpolation by kriging methods in Mohr Basin of Fars province in Iran.
Environmental Monitoring and Assessment, doi: 10.1007/s10661-009-1010-x
Oka T (2005) Analyzing the subsurface geologic structure of the central part of Sapporo City and its
northwest suburb by drilling data of fluid resources, with notes on geological explanation for
six profiles of seimic prospecting performed by the municipal authorities of Sapporo City and
so on. Report of the Geological Survey of Hokkaido, 76, 1-54, Geological Survey of
Hokkaido, In Japanese
(2000) 19881999
1220-32
237
Obara T (1971) On the fluctuation of the ground water level in Toyohira-gawa Fan, Sapporo. Report
of the Geological Survey of Hokkaido, 44, 79-86, Geological Survey of Hokkaido, In
Japanese
(1969) , 39,
Report of the Geological Survey of Hokkaido, 76, 1-54, Geological Survey of
Hokkaido, In Japanese
Ono Y (1990) Alluvial fans in Japan and South Korea, In Alluvial fans: a field approach (ed.
Rachocki AH and Church M). John Wiley and Sons
Ohta A, Tani K, Yamada S and Kaneko S (2007) Mechanical characteristics of undisturbed sandy
gravel including water channel. Proceedings of the forty-second Japan National Conference
of Geotechnical Engineering, Nagoya, Japan, 447-448, the Japanese Geotechniacal Society.
(1974) No.1
Ozaki T, Kishi K, Koma T and Yokota S (1965) Areal Investigation for Ground Water Resources on
the Lower Steams of Toyohira River Basin and Hassamu River Basin, Hokkaido, Bulletin of
the Geological Survey of Japan, 16(1), 1-24, Geological Survey of Japan, In Japanese
Pardo-Igzquiza E and Chica-Olmo M (2004) Estimation of gradients from sparse data by universal
kriging. Water Resources Research, 40(12), doi: 10.1029/2004WR003081
Philip RD and Kitanidis PK (1989) Geostatistical estimation of hydraulic head gradients. Ground
Water, 27(6), 855-865, doi: 10.1111/j.1745-6584.1989.tb01049.x
Proce CJ, Ritzi RW, Dominic DF and Zhenxue D (2004) Modeling multiscale heterogeneity and
aquifer
interconnectivity.
GroundWater,
42(5),
658-670,
doi10.1111/j.1745-6584.2004.tb02720.x
Public Works Research Institute2002 14
238
Sakata Y, Ito K, Isozaki S and Ikeda R (2011) A distribution model of permeability derived from
undisturbed gravelly samples in alluvial fan. Japanese Geotechnical Journal, 6(1), 109119
Sakata Y and Ikeda R (2012a) Quantification of longitudinal river discharge and leakage in an
alluvial Fan by synoptic survey using handheld ADV. Journal of Japan Society of Hydrology
and Water Resources, 25(2), 89-102
Sakata Y and Ikeda R (2012b) Effectiveness of a high resolution model on groundwater simulation
239
159-160
(2011)
(1963)
Scanlon BR, Healy RW and Cook PG (2002) Choosing appropriate techniques for quantifying
groundwater recharge, Hydrogeology Journal, 10, 18-39, doi: 10.1007/s10040-0010176-2
Selley RC (2000) Applied Sedimentology second edition. Academic Press
Sen PK (1968) Estimates of the regression coefficient based on KendallTau. Journal of the
American Statistical Association, 63, 1379-1389
Shepherd RG (1989) Correlations of permeability and grain size. Ground Water, 27, 633-638, doi:
10.1111/j.1745-6584.1989.tb00476.x
Simonds FW and KA Sinclair (2002) Surface Water-Ground Water Interactions Along the Lower
Dungeness River and Vertical Hydraulic Conductivity of Streambed Sediments, Clallam
County, Washington, September 1999-July 2001. Water-Resources Investigations Report,
02-4161, USGS
Singhal BBS and Gupta RP (1999) Applied hydrogeology of fractured rocks. Kluwer academic
publishers
Snow DT (1969) Anisotropic permeability of fractured media. Water Resources Research, 5, doi:
10.1029/WR005i006p01273
240
Soderberg K and Hennet RJC (2007) Uncertainty and Trend Analysis-Radium in Ground Water and
Drinking
Water.
Ground
Water
Monitoring
and
Remediation,
doi:
10.1111/j.1745-6592.2007.00167.x
Song J, Chen X, Cheng C, Wang D, Lackey S and Xu Z (2009) Feasibility of grain-size analysis
methods for determination of vertical hydraulic conductivity of streambeds. Journal of
Hydrology, 375, 428-437, doi: 10.1016/j.jhydrol.2009.06.043
SonTek/YSI (2009) FlowTracker Handheld ADV Technical Manual. SonTek/YSI
Sposito G (1998) Scale Dependence and Scale Invariance in Hydrology. Cambridge university
press
Stonestrom DA and Constantz J (2003) Heat as a tool for studying the movement of ground water
near streams. Circular 1260, USGS
Stone MC and Hotchkiss RH (2007) Evaluating velocity measurement techniques in shallow
streams. Journal of Hydraulic Research, 45, 752-762, doi: 10.1080/00221686.2007.9521813
Takizawa S (2008) Groundwater use and management in urban areas, In Groundwater Management
in Asian Cities Technology and Policy for Sustainability (Ed. Takizawa S), 13-33, Springer
(2009)
20
Tanaka Y, Kudo K, Yosida Y, Nisi K, Aida M and Suzuki H (1990) On the applicability of various
sampling methods to the gravelly ground. Abiko Research Laboratory Reports, U90046,
Central Research Institute of Electric Power Industry
Uchida Y, Sakura Y and Taniguchi M (2001) Shallow subsurface thermal regime in Japan new
concept of subsurface thermal regime-. Journal of geothermal research society of Japan,
23(3), 167-180, In Japanese
Taniguchi M (1993) Evaluation of vertical groundwater fluxes and thermal properties of aquifers
based on transient temperature-depth profiles. Water Resources Research, 29(7), 2021-2026,
doi: 10.1029/93WR00541
Taniguchi M, Shimada J, Tanaka T, Kayane I, Sakura Y, Shimano Y, Dapaah-Siakwan S and
Kawashima S (1999) Disturbances of temperature-depth profiles due to surface climate
241
change and subsurface water flow: 1. An effect of linear increase in surface temperature
caused by global warming and urbanization in the Tokyo metropolitan area, Japan. Water
Resources Research, 35(5), 1507-1517, doi: 10.1029/1999WR900009
Taniguchi M and Uemura T (2005) Effects of urbanization and groundwater flow on the subsurface
temperature in Osaka, Japan. Physics of the earth and planetary interiors, 152, 305313,
doi:10.1016/j.pepi.2005.04.006
Taylor CJ and Alley WM (2001) Ground-water-level monitoring and the importance of long-term
water-level data, U.S. Geological Survey Circular 1217. U.S. Geological Survey
Tebakari T (2010) Example of Flow Regime and Discharge Observation Using the ADCP in the
Uchikawa and the Kurobegawa, 1
Thorne D, Langevin CD and Sukopc MC (2006) Addition of simultaneous heat and solute transport
and variable fluid viscosity to SEAWAT. Computers & Geosciences, 32, 1758-1768, doi:
10.1016/j.cageo.2006.04.005
Todd DK and Mays LW (2005) Groundwater Hydrology 3rd edn. Willey and Sons
Tolman CF (1937) Groundwater. McGraw-Hill
Tonkin MJ and Larson SP (2002) Kriging water levels with a regional-linear and point-logarithmic
drift. Ground Water, doi:10.1111/j.1745-6584.2002.tb02503.x
Tth J (2009) Gravitational Systems of Groundwater Flow Theory, Evaluation, Utilization.
Cambridge University Press
Toyota M, Taira A, Hikita M and Miyahara Y (2010) Field measurements of lake currents in Lake
Suwa by means of ADCP. Japanese Journal of Limnology, 71, 45-52, In Japanese
Turnipseed DP and Sauer VB (2010) Discharge Measurements at Gaging Stations. Techniques and
Methods 3A8, USGS
United Nations Environment Programme (UNEP) (1992) World Atlas of Desertification. London:
Edward Arnold.
Varouchakis EA and Hristopulos DT (2012) Comparison of stochastic and deterministic methods for
mapping groundwater level spatial variability in sparsely monitored basins. Environmental
Monitoring and Assessment, doi: 10.1007/s10661-012-2527-y
242
Vandenbohede A and Lebbe L (2010) Parameter estimation based on vertical heat transport in the
surficial zone. Hydrogeology Journal, 18, 931943, doi: 10.1007/s10040-009-0557-5
Vienken T and Dietrich P (2011) Field evaluation of methods for determining hydraulic conductivity
from grain size data. Journal of Hydrology, 400, 58-71, doi: 10.1016/j.jhydrol.2011.01.022
Vukovic M and Soro A (1992) Determination of hydraulic conductivity of porous media from
grain-size composition. Water Resources Publications
Wackernagel H (2003) Multivariate Geostatistics: An Introduction with Application 2nd edition.
Springer
Wang XS, Jiang XW, Wan L, Song G and Xia Q (2009) Evaluation of depth-dependent porosity and
bulk modulus of a shear using permeability-depth trends. International Journal of Rock
Mechanics & Mining Sciences, 46, 1175-1181, doi: 10.1016/j.ijrmms.2009.02.002
Wen F and Chen X (2006) Evaluation of the impact of groundwater irrigation on streamflow in
Nebraska. Journal of Hydrology, 327, doi: 10.1016/j.jhydrol.2005.12.016
Weight WD (2008) Hydrogeology Field Manual 2nd Edition. McGraw-Hill
Weismann GS and Fogg GE (1999) Multi-scale alluvial fan heterogeneity modeled with transition
probability geostatistics in a sequence stratigraphic framework. Jounral of Hydrology, 226,
48-65
Weissmann GS, Carle SA and Fogg GE (1999) Three-dimensional hydrofacies modeling based on
soil survey analysis and transition probability geostatistics. Water Resources Research, 35(6),
1761-1770
Weissmann GS, Mount JF and Fogg GE (2002) Glacially-driven cycles in accumulation space and
sequence stratigraphy of a stream-dominated alluvial fan, San Joaquin Valley, California,
U.S.A. Journal of Sedimentary Research, 72(2), 240-251
Whiting PJ (2003) Flow Measurement and Characterization. In Tools in Fluvial Geomorphology
(Ed. Kondolf GM and Pigay H), John Wiley and Sons
Wiberg PL and Smith JD (1991) Velocity distribution and bed roughness in high-gradient streams.
Water Resources Research, 27(5), 825838, doi: 10.1029/90WR02770
Winter TC, Harvey JW, Franke OL and Alley WM (1998) Groundwater and Surface Water A Single
Resource. Circular 1139, USGS
243
Woessner WW (2000) Stream and fluvial plain ground water interactions: rescaling hydrogeologic
thought. Ground Water, 38(3), 423-429, doi: 10.1111/j.1745-6584.2000.tb00228.x
Yamaguchi H, Osanai H, Sato O, Futamase K, Obara T, Hayakawa F and Yokoyama E (1965)
Explanatory text of hydrogeological maps of Hokkaido No. 8, Sapporo, special part the
grounds and groundwater of Sapporo environments. Geological survey of Hokkaido, In
Japanese
(1983)
79-93
(1995) 37(6),
4-10
Ye M and Khaleel R (2008) A Markov chain model for characterizing medium heterogeneity and
sediment
layering
structure.
Water
Resources
Research,
44,
W09427,
doi:
10.1029/2008WR006924
Zappa G, Bersezio R, Felletti F and Giudici M (2006) Modeling heterogeneity of gravel-sand,
braided stream, alluvial aquifers at the facies scale. Journal of Hydrology, 325, 134-153, doi:
10.1016/j.jhydrol.2005.10.016
Zhang Y, Gable CW and Person M (2006) Equivalent hydraulic conductivity of an experimental
stratigraphy: implications for basin-scale flow simulations. Water Resources Research, 42,
W05404, doi:10.1029/2005WR004720
Zhang Y, Gable CW and Sheets B (2010) Equivalent hydraulic conductivity of three-dimensional
heterogeneous porous media: an upscaling study based on an experimental stratigraphy.
Journal of Hydrology, 388, 304-320, doi: 10.1016/j.jhydrol.2010.05.009
Zheng C and Bennett GD (1995) Applied Contaminant Transport Modeling: Theory and Practice.
John Wiley & Sons
Zheng C and Wang PP (1999) MT3DMS: A modular threedimensional multispecies transport model
for simulation of advection, dispersion, and chemical reactions of contaminants in
groundwater systems; documentation and users guide, Contract Report SERDP-99-1. US
Army Engineer Research and Development Center
Zlotnik VA, Cardenas MB and Toundykov D (2011) Effects of multiscale snisotropy on basin and
hyporheic
groundwater
Flow.
Ground
Water,
49,
576-583,
doi:
244
10.1111/j.1745-6584.2010.00775.x
245